podium

Runner

50 of 51 criteria have a sourced finding; 49 were also cross-checked by a second agent. Every comparison that includes Runner links here for scope, limitations, atomic answers and sources.

Humans approve the rubric, templates and prototype; they do not review every finding. An agent source cross-check is not a hands-on product test. A missing sourced finding is not evidence that a capability is absent.

Rubric
0.3 · 51 criteria
Source checks
2026-11-16
Agent cross-checked
49 criteria
Comparisons
1

Access and agent choice

ACC-01

Client access

Agent cross-checked partial

Desktop app for macOS and Windows (Linux beta) plus a runner CLI; remote control from other computers is a private beta and an iPhone app is waitlisted; no browser client or editor extension.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service, Remote access through the hosted relay (private beta) and iPhone beta · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: macOS, Windows, Linux (beta) · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • mobile_apps: iPhone app is in private beta with a waitlist; Android not mentioned.
  • browser: No browser client; remote control goes through the desktop app on another computer.
  • mobile_browser: No mobile web client documented.
  • editor_extensions: No editor extension; Open in editor buttons launch external editors.
  • cross_device_continuity: Remote control of another computer sessions via relay is invite-only private beta (mainland China).
9 atomic answers
  • Desktop os [supported] macOS 12+ (Apple Silicon and Intel), Windows 10 1809+ x64, Linux ARM64 and x86_64 in beta (glibc 2.35+). Sources: runner:install, runner:release.
  • Mobile apps [partial] iPhone app is in private beta with a waitlist; Android not mentioned. Sources: runner:site.
  • Browser [unsupported] No browser client; remote control goes through the desktop app on another computer. Sources: runner:remote, runner:srch-team.
  • Mobile browser [unsupported] No mobile web client documented. Sources: runner:remote.
  • Cli [supported] runner CLI controls chats, crews, missions, sessions and messages and works with the app closed. Sources: runner:cli-doc, runner:drive.
  • Editor extensions [unsupported] No editor extension; Open in editor buttons launch external editors. Sources: runner:chats.
  • Task actions [supported] Start chats and missions, answer feed questions, stop/resume/restart slots, review files and diffs, from app or CLI. Sources: runner:missions, runner:cli-doc, runner:chats.
  • Notifications [supported] Desktop notifications for agent needs-input, finished turn, failure and optional mission events. Sources: runner:readme, runner:chats.
  • Cross device continuity [partial] Remote control of another computer sessions via relay is invite-only private beta (mainland China). Sources: runner:remote.

Sources: runner:install, runner:readme, runner:release, runner:site, runner:remote, runner:srch-team, runner:cli-doc, runner:drive, runner:chats, runner:missions

Agent cross-checked partial

Launches separately installed agent CLIs in real terminals: Codex, Claude Code, Antigravity CLI, Copilot CLI, pi, Cursor Agent, Grok Build, OpenCode v2 and opt-in TRAE; custom launch flags and custom role commands.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • model_providers: Models discovered from each CLI; provider choice stays inside the agent CLI.
  • custom_adapters: Roles can use a custom command and every agent accepts custom flags; no generic adapter API beyond built-in runtimes (TRAE is opt-in).
5 atomic answers
  • Harnesses [supported] Nine runtimes with per-agent capability table; Claude Code and Codex are primary. Sources: runner:agents-doc.
  • Model providers [partial] Models discovered from each CLI; provider choice stays inside the agent CLI. Sources: runner:agents-doc.
  • Custom adapters [partial] Roles can use a custom command and every agent accepts custom flags; no generic adapter API beyond built-in runtimes (TRAE is opt-in). Sources: runner:agents-doc.
  • Session integration [supported] Hosts each CLI in a real PTY and reads state through agent hooks, OSC 7501 terminal status or terminal activity; no embedded harness. Sources: runner:agents-doc, runner:arch.
  • Mixed harness operation [supported] A crew can mix agents per slot (e.g. Codex coder with Claude Code reviewer). Sources: runner:crews.

Sources: runner:agents-doc, runner:readme, runner:arch, runner:crews

Agent cross-checked partial

Each agent CLI uses its own login (subscription or API); Runner adds multiple Claude Code accounts and a usage indicator; no vendor credits or Runner-billed models.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • api_accounts: Configured in the agent CLI (pi and OpenCode need a provider configured there).
  • vendor_credits: No Runner-provided model credits.
  • local_models: Possible through agents that support them (pi, OpenCode); Runner has no model runtime.
  • provider_restrictions: Provider terms for driving CLIs are third-party documents not stated in the repository.
7 atomic answers
  • Agent subscriptions [supported] Agent CLI provides the model connection with its own account; Claude Code, Codex, Copilot subscriptions work through their native sign-in. Sources: runner:install, runner:agents-doc.
  • Api accounts [partial] Configured in the agent CLI (pi and OpenCode need a provider configured there). Sources: runner:agents-doc.
  • Vendor credits [unsupported] No Runner-provided model credits. Sources: runner:srch-cost.
  • Local models [partial] Possible through agents that support them (pi, OpenCode); Runner has no model runtime. Sources: runner:agents-doc.
  • Account switching [supported] Multiple Claude Code accounts with a Use switcher. Sources: runner:claude-acc.
  • Vendor account requirement [supported] No Runner account is needed; account sign-in is only for remote access. Sources: runner:remote.
  • Provider restrictions [unknown] Provider terms for driving CLIs are third-party documents not stated in the repository.

Sources: runner:install, runner:claude-acc, runner:agents-doc, runner:srch-cost, runner:remote

Agent cross-checked partial

Download, run first-run Setup to detect or install agent CLIs and sign in, pick a project directory and start a chat or mission; works in any folder; Git panels need a repository; no cloud or VPS installer.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • hosted_git: No GitHub integration; agents handle remotes.
  • managed_cloud_setup: No managed cloud execution.
  • byo_vps_onboarding: No remote-host installer; remote access reaches only another computer running Runner (private beta).
7 atomic answers
  • First task setup [supported] Setup page checks agents, installs npm-packaged CLIs, opens a terminal to sign in and prepares the Pair coding crew. Sources: runner:install.
  • No git [supported] Missions start in any directory; Files shows a folder tree without git status. Sources: runner:first-mission, runner:chats.
  • Local git without remote [supported] Changes and Files use local git only. Sources: runner:chats.
  • Hosted git [unsupported] No GitHub integration; agents handle remotes. Sources: runner:srch-pr.
  • Prerequisite guidance [supported] Install guide lists OS versions, Node and Git for Windows needs, signing steps. Sources: runner:install.
  • Managed cloud setup [unsupported] No managed cloud execution. Sources: runner:remote.
  • Byo vps onboarding [unsupported] No remote-host installer; remote access reaches only another computer running Runner (private beta). Sources: runner:remote.

Sources: runner:install, runner:first-mission, runner:chats, runner:srch-pr, runner:remote

Agent cross-checked partial

Full runner CLI over the local daemon socket with JSON output for chats, crews, missions, sessions and messages; no HTTP API, SDK, MCP server or webhooks.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • api: runnerd.sock typed request table used by app and CLI; not a documented public API.
  • sdk: No SDK.
  • mcp_server: Agents drive Runner through the CLI and an installed skill, not MCP.
  • webhooks: No webhooks.
  • authentication: Local socket access as the user; remote CLI uses account sign-in and PIN pairing.
  • stability: Alpha product; no CLI stability policy found.
8 atomic answers
  • Cli [supported] runner CLI with project, role, crew, mission, chat, session, msg, remote and daemon command groups. Sources: runner:cli-doc.
  • Api [partial] runnerd.sock typed request table used by app and CLI; not a documented public API. Sources: runner:arch, runner:process.
  • Sdk [unsupported] No SDK. Sources: runner:srch-team.
  • Mcp server [unsupported] Agents drive Runner through the CLI and an installed skill, not MCP. Sources: runner:drive.
  • Webhooks [unsupported] No webhooks. Sources: runner:srch-sched.
  • Operations [supported] Create, inspect, stop, resume, restart, message, answer and follow feeds. Sources: runner:cli-doc.
  • Authentication [partial] Local socket access as the user; remote CLI uses account sign-in and PIN pairing. Sources: runner:process, runner:remote.
  • Stability [unknown] Alpha product; no CLI stability policy found.

Sources: runner:cli-doc, runner:drive, runner:arch, runner:process, runner:srch-team, runner:srch-sched, runner:remote

Agent cross-checked partial

Settings manage each agent own MCP servers and skills and install a Runner skill; Runner is not itself an MCP client or server.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • mcp_client: Settings list and toggle MCP servers per agent by editing native config; the agent is the MCP client.
  • plugins: No Runner plugin system.
  • custom_tools: Custom launch flags and role commands.
  • authentication: Inherited from each agent config.
  • permission_scope: Toggles change only the touched entry for new sessions.
7 atomic answers
  • Mcp client [partial] Settings list and toggle MCP servers per agent by editing native config; the agent is the MCP client. Sources: runner:mcp-skills.
  • Plugins [unsupported] No Runner plugin system. Sources: runner:srch-team.
  • Custom tools [partial] Custom launch flags and role commands. Sources: runner:agents-doc.
  • Instructions import [supported] Skills catalog view and switch per agent; roles and crew conventions supply instructions. Sources: runner:mcp-skills, runner:crews.
  • Authentication [partial] Inherited from each agent config. Sources: runner:mcp-skills.
  • Permission scope [partial] Toggles change only the touched entry for new sessions. Sources: runner:mcp-skills.
  • Integration owner [supported] Inherited from the underlying agents; Runner edits their config. Sources: runner:mcp-skills.

Sources: runner:mcp-skills, runner:drive, runner:srch-team, runner:agents-doc, runner:crews

Agent cross-checked partial

Native Rust app using GPUI with alacritty_terminal; no Electron or Tauri.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • mobile_implementation: iPhone beta is private; implementation not public.
  • hybrids: No web hybrid in the public repository.
4 atomic answers
  • Desktop implementation [supported] Rust GPUI UI on a GPU-drawn alacritty_terminal grid, with runnerd daemon. Sources: runner:readme, runner:arch.
  • Mobile implementation [unknown] iPhone beta is private; implementation not public. Sources: runner:site.
  • Hybrids [unsupported] No web hybrid in the public repository. Sources: runner:cargo.
  • Implementation evidence [supported] Workspace crates and README acknowledgements. Sources: runner:cargo, runner:readme.

Sources: runner:readme, runner:cargo, runner:arch, runner:site

Planning and context

Agent cross-checked partial

Work starts from a typed goal in a mission or chat; no tracker, import or sync.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • local_creation: A mission with a goal is the closest entity; no task record.
  • external_import: No tracker import.
  • chat_git_triggers: No chat or git triggers.
  • authoritative_tracker: No tracker.
  • stable_ids: Missions and sessions have IDs (ULIDs) usable in CLI.
  • sync: No sync.
  • supported_edits: No tracker edits.
  • external_roundtrips: Not applicable without a tracker.
9 atomic answers

Sources: runner:missions, runner:srch-issue, runner:cli-doc, runner:srch-sched, runner:drive

Agent cross-checked partial

No plan object; missions have a goal and crews have team conventions.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • plans: No plan entity.
  • acceptance_criteria: No acceptance criteria field; goals describe validation.
  • editors: Roles and crews are editable; goals are text.
  • revisions: No revision history.
  • implementation_links: No links to implementation.
  • amendment: Message members mid-run through the feed.
7 atomic answers
  • Plans [unsupported] No plan entity. Sources: runner:srch-issue.
  • Acceptance criteria [unsupported] No acceptance criteria field; goals describe validation. Sources: runner:drive.
  • Editors [partial] Roles and crews are editable; goals are text. Sources: runner:crews.
  • Revisions [unsupported] No revision history. Sources: runner:srch-issue.
  • Amendment [partial] Message members mid-run through the feed. Sources: runner:missions.
  • Restart persistence [supported] Mission feed and sessions persist across restarts. Sources: runner:process, runner:missions.

Sources: runner:missions, runner:crews, runner:srch-issue, runner:drive, runner:process

Agent cross-checked partial

No epics, dependencies or task hierarchy; handoff order is defined in role prompts and team conventions.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • epics: No such feature (bounded search); crew slots are roles, not work items.
  • hierarchy: No such feature (bounded search); crew slots are roles, not work items.
  • blockers: No such feature (bounded search); crew slots are roles, not work items.
  • readiness: No such feature (bounded search); crew slots are roles, not work items.
  • enforcement: No such feature (bounded search); crew slots are roles, not work items.
  • state_propagation: No such feature (bounded search); crew slots are roles, not work items.
  • task_to_epic: No such feature (bounded search); crew slots are roles, not work items.
  • identity_retention: No such feature (bounded search); crew slots are roles, not work items.
  • workflow_constraints: Crew conventions say who acts first; no rigid workflow engine.
9 atomic answers
  • Epics [unsupported] No such feature (bounded search); crew slots are roles, not work items. Sources: runner:srch-issue.
  • Hierarchy [unsupported] No such feature (bounded search); crew slots are roles, not work items. Sources: runner:srch-issue.
  • Blockers [unsupported] No such feature (bounded search); crew slots are roles, not work items. Sources: runner:srch-issue.
  • Readiness [unsupported] No such feature (bounded search); crew slots are roles, not work items. Sources: runner:srch-issue.
  • Enforcement [unsupported] No such feature (bounded search); crew slots are roles, not work items. Sources: runner:srch-issue.
  • State propagation [unsupported] No such feature (bounded search); crew slots are roles, not work items. Sources: runner:srch-issue.
  • Task to epic [unsupported] No such feature (bounded search); crew slots are roles, not work items. Sources: runner:srch-issue.
  • Identity retention [unsupported] No such feature (bounded search); crew slots are roles, not work items. Sources: runner:srch-issue.
  • Workflow constraints [partial] Crew conventions say who acts first; no rigid workflow engine. Sources: runner:crews.

Sources: runner:crews, runner:srch-issue

Agent cross-checked partial

Roles and crew conventions, skills catalog and Runner skill; no managed memory or project docs store.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • project_docs: Files view reads docs; none stored.
  • memory: No managed memory.
  • scope: Per project, per crew, per agent.
  • provenance: None.
  • correction: None.
  • deletion: None.
  • freshness: None.
  • takeover_continuity: Mission feed and resumable sessions persist; single user.
10 atomic answers

Sources: runner:crews, runner:mcp-skills, runner:missions, runner:drive, runner:chats, runner:srch-issue

Agent cross-checked partial

No indexing or retrieval; Files view with finder, highlighted code and read-only previews beside agents.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • code_search: File finder (⇧⌘O) and command palette; no code search engine.
  • indexing: No such feature (bounded search).
  • external_knowledge: No such feature (bounded search).
  • multimodal_inputs: Not established.
  • freshness: No such feature (bounded search).
  • access_scope: Files view limited to the session checkout or chosen folder.
  • api_retrieval: No such feature (bounded search).
  • citations: No such feature (bounded search).
  • conflict_detection: No such feature (bounded search).
  • review_enforcement: No such feature (bounded search).
10 atomic answers
  • Indexing [unsupported] No such feature (bounded search). Sources: runner:srch-issue.
  • External knowledge [unsupported] No such feature (bounded search). Sources: runner:srch-issue.
  • Multimodal inputs [unknown] Not established.
  • Freshness [unsupported] No such feature (bounded search). Sources: runner:srch-issue.
  • Access scope [partial] Files view limited to the session checkout or chosen folder. Sources: runner:chats.
  • Api retrieval [unsupported] No such feature (bounded search). Sources: runner:srch-issue.
  • Citations [unsupported] No such feature (bounded search). Sources: runner:srch-issue.
  • Conflict detection [unsupported] No such feature (bounded search). Sources: runner:srch-issue.
  • Review enforcement [unsupported] No such feature (bounded search). Sources: runner:srch-issue.

Sources: runner:chats, runner:srch-issue

Agent cross-checked partial

No issue management; missions and chats are pinned, moved, archived and grouped under projects.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • create: No issue or backlog feature (bounded search).
  • edit: No issue or backlog feature (bounded search).
  • discuss: No issue or backlog feature (bounded search).
  • prioritize: No issue or backlog feature (bounded search).
  • label: No issue or backlog feature (bounded search).
  • assign: No issue or backlog feature (bounded search).
  • search: Command palette and sidebar search over chats and missions.
  • filter: No issue or backlog feature (bounded search).
  • reopen: No issue or backlog feature (bounded search).
  • project_epic_views: Projects group chats and missions by directory.
  • idea_parking: No issue or backlog feature (bounded search).
  • explicit_dispatch: No issue or backlog feature (bounded search).
  • task_session_separation: No issue or backlog feature (bounded search).
  • metadata: No issue or backlog feature (bounded search).
  • durable_states: Mission state and status persisted; not issue states.
  • external_tracker_requirement: No issue or backlog feature (bounded search).
17 atomic answers
  • Create [unsupported] No issue or backlog feature (bounded search). Sources: runner:srch-issue.
  • Edit [unsupported] No issue or backlog feature (bounded search). Sources: runner:srch-issue.
  • Discuss [unsupported] No issue or backlog feature (bounded search). Sources: runner:srch-issue.
  • Prioritize [unsupported] No issue or backlog feature (bounded search). Sources: runner:srch-issue.
  • Label [unsupported] No issue or backlog feature (bounded search). Sources: runner:srch-issue.
  • Assign [unsupported] No issue or backlog feature (bounded search). Sources: runner:srch-issue.
  • Filter [unsupported] No issue or backlog feature (bounded search). Sources: runner:srch-issue.
  • Archive [supported] Missions archive and restore. Sources: runner:missions.
  • Reopen [unsupported] No issue or backlog feature (bounded search). Sources: runner:srch-issue.
  • Project epic views [partial] Projects group chats and missions by directory. Sources: runner:chats.
  • Idea parking [unsupported] No issue or backlog feature (bounded search). Sources: runner:srch-issue.
  • Explicit dispatch [unsupported] No issue or backlog feature (bounded search). Sources: runner:srch-issue.
  • Task session separation [unsupported] No issue or backlog feature (bounded search). Sources: runner:srch-issue.
  • Metadata [unsupported] No issue or backlog feature (bounded search). Sources: runner:srch-issue.
  • Durable states [partial] Mission state and status persisted; not issue states. Sources: runner:missions.
  • External tracker requirement [unsupported] No issue or backlog feature (bounded search). Sources: runner:srch-issue.

Sources: runner:missions, runner:srch-issue, runner:chats

Agent cross-checked partial

Agents operate missions through the runner CLI, not issues: start, follow, message, answer and read feed; no claims or dependencies.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • shared_ids: Mission and session IDs shared by app and CLI.
  • update: Rename, move, link missions.
  • claim: No work-item operations.
  • assign: No work-item operations.
  • dependencies: No work-item operations.
  • children: No work-item operations.
  • attribution: Feed messages attributed to roster handles.
  • permissions: No work-item operations.
  • conflicts: No work-item operations.
14 atomic answers
  • Shared ids [partial] Mission and session IDs shared by app and CLI. Sources: runner:drive.
  • Read [supported] runner mission show/feed. Sources: runner:cli-doc.
  • Create [supported] runner mission start from an agent. Sources: runner:drive.
  • Update [partial] Rename, move, link missions. Sources: runner:cli-doc.
  • Claim [unsupported] No work-item operations. Sources: runner:srch-issue.
  • Assign [unsupported] No work-item operations. Sources: runner:srch-issue.
  • State changes [supported] Stop, resume, archive. Sources: runner:cli-doc.
  • Dependencies [unsupported] No work-item operations. Sources: runner:srch-issue.
  • Children [unsupported] No work-item operations. Sources: runner:srch-issue.
  • Results [supported] Reports back to the linked chat. Sources: runner:missions.
  • Human visibility [supported] Same missions visible in app. Sources: runner:missions.
  • Attribution [partial] Feed messages attributed to roster handles. Sources: runner:missions.
  • Permissions [unsupported] No work-item operations. Sources: runner:srch-issue.
  • Conflicts [unsupported] No work-item operations. Sources: runner:srch-issue.

Sources: runner:drive, runner:cli-doc, runner:srch-issue, runner:missions

Agent cross-checked unsupported

No living specification feature.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • persistent_requirements: No spec store (bounded search).
  • human_edits: No spec store (bounded search).
  • agent_edits: No spec store (bounded search).
  • revisions: No spec store (bounded search).
  • approvals: No spec store (bounded search).
  • implementation_links: No spec store (bounded search).
  • later_retrieval: No spec store (bounded search).
  • drift_detection: No spec store (bounded search).
  • review_integration: No spec store (bounded search).
9 atomic answers
  • Persistent requirements [unsupported] No spec store (bounded search). Sources: runner:srch-issue.
  • Human edits [unsupported] No spec store (bounded search). Sources: runner:srch-issue.
  • Agent edits [unsupported] No spec store (bounded search). Sources: runner:srch-issue.
  • Revisions [unsupported] No spec store (bounded search). Sources: runner:srch-issue.
  • Approvals [unsupported] No spec store (bounded search). Sources: runner:srch-issue.
  • Later retrieval [unsupported] No spec store (bounded search). Sources: runner:srch-issue.
  • Drift detection [unsupported] No spec store (bounded search). Sources: runner:srch-issue.
  • Review integration [unsupported] No spec store (bounded search). Sources: runner:srch-issue.

Sources: runner:srch-issue

Coordination and automation

COORD-01

Parallel work

Agent cross-checked partial

Many sessions, tabs and missions run concurrently under runnerd; split panes and a sidebar show working, idle and needs-you states; no documented cap.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • quotas: No quota documented or found.
  • observed_concurrency: No benchmark published; ADEMark does not list Runner.
5 atomic answers
  • Concurrent sessions [supported] Every chat and mission member is its own PTY session; split tabs support many panes. Sources: runner:sessions, runner:readme.
  • Concurrent tasks [supported] Several missions can run at once. Sources: runner:missions.
  • Quotas [unsupported] No quota documented or found. Sources: runner:srch-limits.
  • Observed concurrency [unknown] No benchmark published; ADEMark does not list Runner.
  • Status visibility [supported] Working/idle/needs-you indicators, mission states and unread dots in the sidebar. Sources: runner:sessions, runner:missions.

Sources: runner:sessions, runner:process, runner:missions, runner:readme, runner:srch-limits

Agent cross-checked partial

Crew slots assign roles to agents; messages route by handle; no claims or queues.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • human_assignment: No assignment to humans.
  • claims: No claims.
  • queues: No queue.
  • duplicate_prevention: No duplicate-work prevention; goal text must avoid overlapping edits.
  • ownership_enforcement: No enforcement.
7 atomic answers
  • Human assignment [unsupported] No assignment to humans. Sources: runner:srch-team.
  • Agent assignment [supported] Slots bind roles/agents; goal and conventions say who acts. Sources: runner:crews.
  • Claims [unsupported] No claims. Sources: runner:srch-issue.
  • Queues [unsupported] No queue. Sources: runner:srch-issue.
  • Routing [supported] Handle addressing and @mentions deliver inbox notices to recipients. Sources: runner:missions.
  • Duplicate prevention [unsupported] No duplicate-work prevention; goal text must avoid overlapping edits. Sources: runner:build-crew.
  • Ownership enforcement [unsupported] No enforcement. Sources: runner:build-crew.

Sources: runner:crews, runner:missions, runner:srch-team, runner:srch-issue, runner:build-crew

Agent cross-checked supported

Mission members message each other by handle through an inbox in a persistent feed; an agent in a chat can delegate to a crew via the runner CLI skill and receives reports; per-slot agent/model/effort.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
12 atomic answers
  • Roles [supported] Role templates and slot handles (coder, reviewer). Sources: runner:crews.
  • Model harness effort [supported] Per-slot agent, model, effort and Codex speed overrides. Sources: runner:crews.
  • Messages [supported] Targeted and broadcast messages, questions to the human. Sources: runner:missions.
  • Results return [supported] Handoffs in the feed and state reports to the linked chat. Sources: runner:missions.
  • Cross harness [supported] Slots can use different agents. Sources: runner:crews.
  • Orchestration owner [supported] Product orchestration: daemon delivers inbox notices and reports; coordinator role can be a third-party agent using the shipped skill. Sources: runner:drive, runner:missions.
  • Activation [supported] Always available; skill toggle in Settings → General → Command line. Sources: runner:drive.
  • Authentication [supported] Local socket; agents use their own accounts. Sources: runner:process.
  • Live progress [supported] Live terminals per member and the feed. Sources: runner:missions.
  • Retained results [supported] Feed persists in an event log; archived missions keep their feed. Sources: runner:missions, runner:arch.
  • Followup [supported] Resume stopped sessions and send follow-up messages after completion. Sources: runner:missions.
  • Invocation [supported] runner mission start from chat or shell. Sources: runner:drive.

Sources: runner:missions, runner:drive, runner:crews, runner:process, runner:arch

Agent cross-checked unsupported

No schedules or event triggers.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • schedules: No scheduler or trigger code (bounded search).
  • events: No scheduler or trigger code (bounded search).
  • recurring_tasks: No scheduler or trigger code (bounded search).
  • completion_conditions: No scheduler or trigger code (bounded search).
  • scheduler_location: No scheduler or trigger code (bounded search).
  • overlaps: No scheduler or trigger code (bounded search).
  • retries: No scheduler or trigger code (bounded search).
  • client_independence: No scheduler or trigger code (bounded search).
8 atomic answers
  • Schedules [unsupported] No scheduler or trigger code (bounded search). Sources: runner:srch-sched.
  • Events [unsupported] No scheduler or trigger code (bounded search). Sources: runner:srch-sched.
  • Recurring tasks [unsupported] No scheduler or trigger code (bounded search). Sources: runner:srch-sched.
  • Completion conditions [unsupported] No scheduler or trigger code (bounded search). Sources: runner:srch-sched.
  • Scheduler location [unsupported] No scheduler or trigger code (bounded search). Sources: runner:srch-sched.
  • Overlaps [unsupported] No scheduler or trigger code (bounded search). Sources: runner:srch-sched.
  • Retries [unsupported] No scheduler or trigger code (bounded search). Sources: runner:srch-sched.
  • Client independence [unsupported] No scheduler or trigger code (bounded search). Sources: runner:srch-sched.

Sources: runner:srch-sched

Agent cross-checked partial

Stop, resume and restart per slot or mission; steer by messages or typing in terminals; needs-you reports and desktop notifications; interrupted state after restarts.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • pause: Stop halts a slot; no pause that preserves the process.
  • escalation: Agents ask the human via the feed.
10 atomic answers
  • Pause [partial] Stop halts a slot; no pause that preserves the process. Sources: runner:missions.
  • Stop [supported] Stop slot or mission; history kept; repository changes never reverted. Sources: runner:missions.
  • Resume [supported] Resume saved conversations; Restart starts fresh. Sources: runner:missions.
  • Steer [supported] Feed messages and direct terminal input. Sources: runner:missions.
  • Notifications [supported] Desktop notifications, unread dots, linked-chat reports. Sources: runner:chats, runner:missions.
  • Pending decisions [supported] Questions on feed cards; Needs you state. Sources: runner:missions.
  • Escalation [partial] Agents ask the human via the feed. Sources: runner:missions.
  • Visible errors [supported] Failed and Interrupted mission states and per-slot errors. Sources: runner:missions.
  • Chat terminal parity [supported] Chats are terminals; the same indicators apply. Sources: runner:sessions.
  • Retained history [supported] Feed persistent log. Sources: runner:missions, runner:arch.

Sources: runner:missions, runner:sessions, runner:chats, runner:arch

Agent cross-checked partial

Sidebar tree of projects, chats and missions with per-item state and unread markers; no board of epics or dependencies.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • fleet_fixture: No demo fleet; docs show screenshots.
  • parent_child: Missions group under the chat that started them.
  • dependencies: None.
  • history: Feeds and archived missions.
  • workload: Usage indicator for plan limits.
  • navigation_errors: Command palette and sidebar navigation.
  • fixture_mode: Not established.
8 atomic answers
  • Fleet fixture [unknown] No demo fleet; docs show screenshots.
  • Parent child [partial] Missions group under the chat that started them. Sources: runner:missions.
  • Dependencies [unsupported] None. Sources: runner:srch-issue.
  • Attention [supported] Needs you / Failed / Interrupted states and notifications. Sources: runner:missions.
  • History [partial] Feeds and archived missions. Sources: runner:missions.
  • Workload [partial] Usage indicator for plan limits. Sources: runner:arch.
  • Navigation errors [partial] Command palette and sidebar navigation. Sources: runner:chats.
  • Fixture mode [unknown] Not established.

Sources: runner:missions, runner:chats, runner:srch-issue, runner:arch

Agent cross-checked partial

Crews are reusable presets with per-slot configuration; missions can start empty and add roles; configuration is by role prompts, not a natural-language builder.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • growing_epic: Empty missions can add or remove members while running.
  • natural_language: Role instructions are free text.
  • retained_relationships: Roster captured in the mission.
  • auth_friction: Needs each agent signed in.
10 atomic answers
  • Quick task [supported] A chat or one-line mission. Sources: runner:missions.
  • Growing epic [partial] Empty missions can add or remove members while running. Sources: runner:missions.
  • Presets [supported] Saved crews and the Pair coding starter crew. Sources: runner:crews.
  • Natural language [partial] Role instructions are free text. Sources: runner:crews.
  • Agent configuration [supported] Per-slot agent/model/effort. Sources: runner:crews.
  • In run changes [supported] Add/remove role mid-mission; change messages. Sources: runner:missions.
  • Live roles [supported] Roster and session cards show roles. Sources: runner:missions.
  • Retained relationships [partial] Roster captured in the mission. Sources: runner:missions.
  • Discoverability [supported] Launcher, palette and docs. Sources: runner:missions.
  • Auth friction [partial] Needs each agent signed in. Sources: runner:install.

Sources: runner:crews, runner:missions, runner:install

Execution and reliability

Sourced partial

Local execution on the user computer; sessions live in runnerd and survive app quit; runnerd and the CLI can also run on a Linux host reached over SSH (optional systemd user service) though no guide presents server execution; remote control of another computer is an invite-only private beta; no vendor cloud.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: local · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
Not separately cross-checked
  • byo_remote: Docs describe the Linux runnerd systemd user service and using the CLI over SSH, so the daemon can run on a user-controlled host; remote and server execution is not presented as a supported workflow and relay-based remote access is private beta.
  • self_hosted: The relay service is Runner hosted and closed.
  • managed_cloud: No managed cloud execution.
  • hybrid: Control another computer from this one via relay (beta).
  • provisioning: None.
  • region: Remote access currently available only in mainland China.
  • expiry: No expiry.
  • other_device_continuation: Another computer can attach via the private-beta relay.
12 atomic answers
  • Local [supported] Sessions run in runnerd on the user machine. Sources: runner:arch, runner:process.
  • Byo remote [partial] Docs describe the Linux runnerd systemd user service and using the CLI over SSH, so the daemon can run on a user-controlled host; remote and server execution is not presented as a supported workflow and relay-based remote access is private beta. Sources: runner:remote, runner:linux-service.
  • Self hosted [unsupported] The relay service is Runner hosted and closed. Sources: runner:remote.
  • Managed cloud [unsupported] No managed cloud execution. Sources: runner:remote, runner:srch-team.
  • Hybrid [partial] Control another computer from this one via relay (beta). Sources: runner:remote.
  • Provisioning [unsupported] None. Sources: runner:remote.
  • Payer [supported] User machine and agent subscriptions. Sources: runner:install.
  • Resources [supported] Host machine resources. Sources: runner:install.
  • Region [partial] Remote access currently available only in mainland China. Sources: runner:remote.
  • Expiry [unsupported] No expiry. Sources: runner:process.
  • Client disconnect [supported] Quit or crash of the app leaves daemon sessions running; reopen to the same terminals. Sources: runner:process, runner:readme.
  • Other device continuation [partial] Another computer can attach via the private-beta relay. Sources: runner:remote.

Sources: runner:arch, runner:remote, runner:install, runner:process, runner:srch-team, runner:readme, runner:linux-service

Agent cross-checked partial

Members share the given directory; Runner creates no branches, worktrees or copies; Files can switch between existing worktrees; no container, VM or OS sandbox of its own.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • branch: Does not create branches; default crew conventions suggest a feature branch.
  • worktree: Docs state Runner does not isolate with worktrees; user must create one before launch.
  • container: None.
  • vm: None.
  • os_sandbox: Missions use agent bypass/broad-access modes.
  • filesystem: No filesystem boundary; agents run as the user.
  • process: Separate PTY per session.
  • network: None.
  • secrets: Relay uses PIN-based pairing and OS-stored keys; agent secrets stay in agent configs.
  • code_separation: Shared directory.
  • security_boundary: No containment claim; docs advise trusted folders.
12 atomic answers

Sources: runner:missions, runner:build-crew, runner:first-mission, runner:crews, runner:srch-team, runner:agents-doc, runner:sessions, runner:remote

Agent cross-checked partial

runnerd keeps sessions alive across app quit or crash; stopped agents resume their conversations; interrupted missions are flagged after a daemon restart; a daemon crash ends PTYs.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • worker_crash: Session crash marks it stopped; daemon crash closes PTYs on macOS and kills sessions.
  • expiry: None.
  • state_reconstruction: Interrupted state recorded; chat reports on restart.
  • context_loss: Interrupted turn is lost; conversation resume depends on the agent.
  • duplicate_actions: Not stated.
  • manual_repair: Daemon troubleshooting in docs; use Resume/Restart.
  • multi_day_continuity: Sessions persist while the machine runs; resume after restart.
12 atomic answers
  • Client disconnect [supported] App quit/crash leaves sessions running. Sources: runner:process.
  • Worker crash [partial] Session crash marks it stopped; daemon crash closes PTYs on macOS and kills sessions. Sources: runner:process, runner:arch.
  • Service restart [supported] Members resume at next start; mission shows Interrupted when a turn was lost. Sources: runner:missions.
  • Expiry [unsupported] None. Sources: runner:process.
  • Transcript reload [supported] Feed event log and agent conversations resume. Sources: runner:missions, runner:arch.
  • Process reconnect [supported] Reopen the app to the same live terminals. Sources: runner:readme.
  • Task resume [supported] Resume saved conversation per session where supported. Sources: runner:agents-doc, runner:missions.
  • State reconstruction [partial] Interrupted state recorded; chat reports on restart. Sources: runner:missions.
  • Context loss [partial] Interrupted turn is lost; conversation resume depends on the agent. Sources: runner:missions, runner:agents-doc.
  • Duplicate actions [unknown] Not stated.
  • Manual repair [partial] Daemon troubleshooting in docs; use Resume/Restart. Sources: runner:missions.
  • Multi day continuity [partial] Sessions persist while the machine runs; resume after restart. Sources: runner:process.

Sources: runner:process, runner:missions, runner:arch, runner:readme, runner:agents-doc

Agent cross-checked unsupported

No dependency, cache, service or port management; terminals show what agents do.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • dependencies: No environment management (bounded search).
  • caches: No environment management (bounded search).
  • snapshots: No environment management (bounded search).
  • services: No environment management (bounded search).
  • ports: No environment management (bounded search).
  • secrets: No environment management (bounded search).
  • duplicate_installs: No environment management (bounded search).
  • collisions: No environment management (bounded search).
  • remote_setup: No environment management (bounded search).
  • automatic_repair: No environment management (bounded search).
10 atomic answers
  • Dependencies [unsupported] No environment management (bounded search). Sources: runner:srch-issue.
  • Caches [unsupported] No environment management (bounded search). Sources: runner:srch-issue.
  • Snapshots [unsupported] No environment management (bounded search). Sources: runner:srch-issue.
  • Services [unsupported] No environment management (bounded search). Sources: runner:srch-issue.
  • Ports [unsupported] No environment management (bounded search). Sources: runner:srch-issue.
  • Secrets [unsupported] No environment management (bounded search). Sources: runner:srch-issue.
  • Duplicate installs [unsupported] No environment management (bounded search). Sources: runner:srch-issue.
  • Collisions [unsupported] No environment management (bounded search). Sources: runner:srch-issue.
  • Remote setup [unsupported] No environment management (bounded search). Sources: runner:srch-issue.
  • Automatic repair [unsupported] No environment management (bounded search). Sources: runner:srch-issue.

Sources: runner:chats, runner:srch-issue

Agent cross-checked partial

No agent time, turn, token or spend limits are enforced; usage indicator shows provider plan-limit percentages.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • time_limits: No limit enforcement (bounded search).
  • turn_limits: No limit enforcement (bounded search).
  • token_limits: No limit enforcement (bounded search).
  • spend_limits: No limit enforcement (bounded search).
  • timeouts: Only operational timeouts for Runner internals.
  • retries: No limit enforcement (bounded search).
  • provider_failures: No limit enforcement (bounded search).
  • enforcement: Plan limits are shown, not enforced.
8 atomic answers
  • Time limits [unsupported] No limit enforcement (bounded search). Sources: runner:srch-limits.
  • Turn limits [unsupported] No limit enforcement (bounded search). Sources: runner:srch-limits.
  • Token limits [unsupported] No limit enforcement (bounded search). Sources: runner:srch-limits.
  • Spend limits [unsupported] No limit enforcement (bounded search). Sources: runner:srch-limits.
  • Timeouts [partial] Only operational timeouts for Runner internals. Sources: runner:srch-limits.
  • Retries [unsupported] No limit enforcement (bounded search). Sources: runner:srch-limits.
  • Provider failures [unsupported] No limit enforcement (bounded search). Sources: runner:srch-limits.
  • Enforcement [unsupported] Plan limits are shown, not enforced. Sources: runner:arch.

Sources: runner:srch-limits, runner:arch

Agent cross-checked partial

Terminal scrollback, mission feed event log, session states and usage-limit indicators; no traces or cost statistics.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • traces: None.
  • failure_reasons: Failed/Interrupted states and per-session errors.
  • usage_export: None found.
  • task_stats: None.
  • agent_stats: Per-agent plan usage indicator.
  • project_stats: None.
  • time_stats: Weekly and 5-hour windows.
  • reconnect_consistency: Hook-based state with bridge authority; interrupted marking.
  • performance_method: No benchmarks; ADEMark does not list Runner.
  • responsiveness: No figures.
13 atomic answers
  • Logs [supported] runnerd log folder and feed event log. Sources: runner:arch.
  • Traces [unsupported] None. Sources: runner:srch-tele.
  • Failure reasons [partial] Failed/Interrupted states and per-session errors. Sources: runner:missions.
  • Usage export [unsupported] None found. Sources: runner:srch-cost.
  • Task stats [unsupported] None. Sources: runner:srch-cost.
  • Agent stats [partial] Per-agent plan usage indicator. Sources: runner:arch.
  • Project stats [unsupported] None. Sources: runner:srch-cost.
  • Time stats [partial] Weekly and 5-hour windows. Sources: runner:claude-acc.
  • Failure status [supported] Failed state. Sources: runner:missions.
  • Approval status [supported] Needs you with approval waits. Sources: runner:agents-doc.
  • Reconnect consistency [partial] Hook-based state with bridge authority; interrupted marking. Sources: runner:agents-doc, runner:missions.
  • Performance method [unknown] No benchmarks; ADEMark does not list Runner.
  • Responsiveness [unknown] No figures.

Sources: runner:arch, runner:missions, runner:sessions, runner:srch-tele, runner:srch-cost, runner:claude-acc, runner:agents-doc

Agent cross-checked partial

No dollar cost or token accounting; plan-limit usage percentages for Claude Code, Codex and Antigravity CLI.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • task_cost: No cost accounting (bounded search).
  • session_cost: No cost accounting (bounded search).
  • child_cost: No cost accounting (bounded search).
  • epic_cost: No cost accounting (bounded search).
  • harness_aggregation: No aggregation across harnesses.
  • billed_vs_estimated: No cost accounting (bounded search).
  • tokens: No cost accounting (bounded search).
  • subscription_allowance: Shows subscription limit windows (% used) for supported agents.
  • provider_reconciliation: No cost accounting (bounded search).
  • compute_cost: No cost accounting (bounded search).
  • refresh_delay: Configurable check interval.
  • alerts: No cost accounting (bounded search).
  • export: No cost accounting (bounded search).
  • drilldown: No cost accounting (bounded search).
14 atomic answers
  • Task cost [unsupported] No cost accounting (bounded search). Sources: runner:srch-cost.
  • Session cost [unsupported] No cost accounting (bounded search). Sources: runner:srch-cost.
  • Child cost [unsupported] No cost accounting (bounded search). Sources: runner:srch-cost.
  • Epic cost [unsupported] No cost accounting (bounded search). Sources: runner:srch-cost.
  • Harness aggregation [unsupported] No aggregation across harnesses. Sources: runner:srch-cost.
  • Billed vs estimated [unsupported] No cost accounting (bounded search). Sources: runner:srch-cost.
  • Tokens [unsupported] No cost accounting (bounded search). Sources: runner:srch-cost.
  • Subscription allowance [partial] Shows subscription limit windows (% used) for supported agents. Sources: runner:arch, runner:claude-acc.
  • Provider reconciliation [unsupported] No cost accounting (bounded search). Sources: runner:srch-cost.
  • Compute cost [unsupported] No cost accounting (bounded search). Sources: runner:srch-cost.
  • Refresh delay [partial] Configurable check interval. Sources: runner:claude-acc.
  • Alerts [unsupported] No cost accounting (bounded search). Sources: runner:srch-cost.
  • Export [unsupported] No cost accounting (bounded search). Sources: runner:srch-cost.
  • Drilldown [unsupported] No cost accounting (bounded search). Sources: runner:srch-cost.

Sources: runner:arch, runner:srch-cost, runner:claude-acc

Agent cross-checked unsupported

No queues, resource limits or deduplication of checks.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • concurrency: No such feature (bounded search).
  • queues: No such feature (bounded search).
  • cpu_ram: No such feature (bounded search).
  • backpressure: No such feature (bounded search).
  • host_awareness: No such feature (bounded search).
  • check_reuse: No such feature (bounded search).
  • check_deduplication: No such feature (bounded search).
  • bounded_tests: No such feature (bounded search).
  • cancellation: No such feature (bounded search).
  • fairness: No such feature (bounded search).
  • recovery: No such feature (bounded search).
  • shared_limits: No such feature (bounded search).
12 atomic answers

Sources: runner:srch-limits

Human collaboration and control

Agent cross-checked unsupported

Single-user local app; remote control by the same account on another computer; no sharing or co-prompting.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • visibility: No multi-user code (bounded search); remote access is account-bound.
  • discussion: No multi-user code (bounded search); remote access is account-bound.
  • comments: No multi-user code (bounded search); remote access is account-bound.
  • presence: No multi-user code (bounded search); remote access is account-bound.
  • coprompting: No multi-user code (bounded search); remote access is account-bound.
  • ordering: No multi-user code (bounded search); remote access is account-bound.
  • conflicts: No multi-user code (bounded search); remote access is account-bound.
  • attribution: No multi-user code (bounded search); remote access is account-bound.
  • permissions: No multi-user code (bounded search); remote access is account-bound.
  • plan_restrictions: No multi-user code (bounded search); remote access is account-bound.
  • owner_online: No multi-user code (bounded search); remote access is account-bound.
11 atomic answers

Sources: runner:remote, runner:srch-team

Agent cross-checked unsupported

No roles, organizations or SSO.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • organization_roles: No roles, membership or SSO (bounded search).
  • project_roles: No roles, membership or SSO (bounded search).
  • task_roles: No roles, membership or SSO (bounded search).
  • membership: No roles, membership or SSO (bounded search).
  • sso: No roles, membership or SSO (bounded search).
  • provisioning: No roles, membership or SSO (bounded search).
  • unauthorized_access: No roles, membership or SSO (bounded search).
  • revocation: No roles, membership or SSO (bounded search).
  • plan_scope: No roles, membership or SSO (bounded search).
9 atomic answers
  • Organization roles [unsupported] No roles, membership or SSO (bounded search). Sources: runner:srch-team.
  • Project roles [unsupported] No roles, membership or SSO (bounded search). Sources: runner:srch-team.
  • Task roles [unsupported] No roles, membership or SSO (bounded search). Sources: runner:srch-team.
  • Membership [unsupported] No roles, membership or SSO (bounded search). Sources: runner:srch-team.
  • Sso [unsupported] No roles, membership or SSO (bounded search). Sources: runner:srch-team.
  • Provisioning [unsupported] No roles, membership or SSO (bounded search). Sources: runner:srch-team.
  • Unauthorized access [unsupported] No roles, membership or SSO (bounded search). Sources: runner:srch-team.
  • Revocation [unsupported] No roles, membership or SSO (bounded search). Sources: runner:srch-team.
  • Plan scope [unsupported] No roles, membership or SSO (bounded search). Sources: runner:srch-team.

Sources: runner:srch-team

Agent cross-checked partial

Agent permission behavior is the agent own: chats default to the agent configuration, Bypass toggles are opt-in, and missions always use broad access.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • command: Chats follow each agent approval setting unless Bypass is on; Needs you reports waits.
  • tool: Same.
  • network: None.
  • secret: None.
  • delivery: None.
  • approver_identity: Single user.
  • lifetime: None.
  • revoke: Pairing can be revoked for remote devices.
  • tunnels: No tunnels besides the relay.
  • safer_modes: Leave Bypass off for chats; pi uses trusted workspace.
  • enforcement: Not product-enforced; goal text sets limits.
14 atomic answers

Sources: runner:agents-doc, runner:missions, runner:srch-pr, runner:srch-team, runner:remote

Agent cross-checked partial

Persistent mission feed and session history locally; no attribution of humans, audit or export.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • actions: Feed records messages, questions and lifecycle events.
  • approvals: Approvals not recorded by Runner.
  • edits: None.
  • decisions: Answers on feed cards.
  • cross_user_persistence: Single user.
  • export: No export.
  • mutability: Not stated.
  • retention: Not stated.
  • integrity: No integrity guarantee; no fsync by design.
11 atomic answers
  • Actions [partial] Feed records messages, questions and lifecycle events. Sources: runner:missions.
  • Approvals [unsupported] Approvals not recorded by Runner. Sources: runner:agents-doc.
  • Edits [unsupported] None. Sources: runner:chats.
  • Discussion [supported] Feed. Sources: runner:missions.
  • Decisions [partial] Answers on feed cards. Sources: runner:missions.
  • Cross user persistence [unsupported] Single user. Sources: runner:srch-team.
  • Cross session persistence [supported] SQLite and event log. Sources: runner:arch.
  • Export [unsupported] No export. Sources: runner:srch-issue.
  • Mutability [unknown] Not stated.
  • Retention [unknown] Not stated.
  • Integrity [unsupported] No integrity guarantee; no fsync by design. Sources: runner:arch.

Sources: runner:missions, runner:arch, runner:agents-doc, runner:chats, runner:srch-team, runner:srch-issue

Agent cross-checked partial

Files view with git status and a read-only Changes view with per-file diffs and mark-viewed beside agents; no comments, request-changes or review queue.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • pending_reviews: Changes lists files with uncommitted changes and commits ahead.
  • evidence_navigation: Open files, terminal links ⌘-click.
  • comments: No comments.
  • request_changes: Message the reviewer/coder in the feed.
  • shipping_guidance: None.
  • effort: Not published.
  • missed_evidence: Mark viewed dims files.
  • narrow_screen: Desktop only.
  • roles: Reviewer role in the default crew.
13 atomic answers
  • Pending reviews [partial] Changes lists files with uncommitted changes and commits ahead. Sources: runner:chats.
  • Change understanding [supported] Per-file diff, line counts and rename info. Sources: runner:chats.
  • Evidence navigation [partial] Open files, terminal links ⌘-click. Sources: runner:chats, runner:sessions.
  • Comments [unsupported] No comments. Sources: runner:srch-pr.
  • Request changes [partial] Message the reviewer/coder in the feed. Sources: runner:missions.
  • Shipping guidance [unsupported] None. Sources: runner:srch-pr.
  • Effort [unknown] Not published.
  • Missed evidence [partial] Mark viewed dims files. Sources: runner:chats.
  • Keyboard [supported] Shortcut-driven (Space, C, arrows) and a keyboard shortcuts reference. Sources: runner:chats.
  • Narrow screen [unknown] Desktop only.
  • Context switching [supported] Split tabs and side panel. Sources: runner:readme, runner:chats.
  • Roles [partial] Reviewer role in the default crew. Sources: runner:crews.
  • Retained context [supported] Panel state persisted per mission. Sources: runner:missions.

Sources: runner:chats, runner:sessions, runner:srch-pr, runner:missions, runner:readme, runner:crews

Agent cross-checked partial

No colleague takeover; the same account can take a session over from another of its computers in the private beta.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • ongoing_takeover: Another computer on the same account can take terminal control (private beta).
  • same_task: Same missions and chats of the host.
  • decisions: Feed answers.
  • artifacts: None.
  • permissions: PIN pairing per device; revoke devices.
  • owner_disconnect: Host must be awake and signed in.
  • ownership: No transfer.
  • attribution: No attribution.
  • context_loss: Not stated.
  • harness_transfer: No cross-harness transfer.
12 atomic answers
  • Ongoing takeover [partial] Another computer on the same account can take terminal control (private beta). Sources: runner:remote.
  • Same task [partial] Same missions and chats of the host. Sources: runner:remote.
  • Session history [supported] Remote opens live terminals and feed. Sources: runner:remote.
  • Decisions [partial] Feed answers. Sources: runner:remote.
  • Artifacts [unsupported] None. Sources: runner:srch-issue.
  • Permissions [partial] PIN pairing per device; revoke devices. Sources: runner:remote.
  • Runtime context [supported] Execution stays on the host. Sources: runner:remote.
  • Owner disconnect [unsupported] Host must be awake and signed in. Sources: runner:remote.
  • Ownership [unsupported] No transfer. Sources: runner:srch-team.
  • Attribution [unsupported] No attribution. Sources: runner:srch-team.
  • Context loss [unknown] Not stated.
  • Harness transfer [unsupported] No cross-harness transfer. Sources: runner:agents-doc.

Sources: runner:remote, runner:srch-issue, runner:srch-team, runner:agents-doc

Verification and delivery

Agent cross-checked partial

Terminal output, mission feed and read-only file and diff views; no artifact gallery.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • documents: Markdown rendered in Files.
  • screenshots: None.
  • recordings: None.
  • previews: No preview server integration.
  • test_outputs: In terminals only.
  • reviewer_outputs: Reviewer member messages in the feed.
  • gallery: None.
  • task_links: None.
  • revision_links: None.
  • prompt_leakage: Not assessed.
13 atomic answers

Sources: runner:chats, runner:missions, runner:srch-issue, runner:sessions

Agent cross-checked partial

No gates; conventions can require checks but are instructions to the agents.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • tests: No gate in the product (bounded search).
  • lint: No gate in the product (bounded search).
  • types: No gate in the product (bounded search).
  • ci: No gate in the product (bounded search).
  • independent_review: Default crew has a reviewer agent role, as an instruction.
  • spec_checks: No gate in the product (bounded search).
  • required_checks: No gate in the product (bounded search).
  • failure_blocking: No gate in the product (bounded search).
  • requirement_conflicts: No gate in the product (bounded search).
  • configured_policy: Team conventions text.
  • enforced_policy: No gate in the product (bounded search).
11 atomic answers
  • Tests [unsupported] No gate in the product (bounded search). Sources: runner:srch-pr.
  • Lint [unsupported] No gate in the product (bounded search). Sources: runner:srch-pr.
  • Types [unsupported] No gate in the product (bounded search). Sources: runner:srch-pr.
  • Ci [unsupported] No gate in the product (bounded search). Sources: runner:srch-pr.
  • Independent review [partial] Default crew has a reviewer agent role, as an instruction. Sources: runner:crews.
  • Spec checks [unsupported] No gate in the product (bounded search). Sources: runner:srch-pr.
  • Required checks [unsupported] No gate in the product (bounded search). Sources: runner:srch-pr.
  • Failure blocking [unsupported] No gate in the product (bounded search). Sources: runner:srch-pr.
  • Requirement conflicts [unsupported] No gate in the product (bounded search). Sources: runner:srch-pr.
  • Configured policy [partial] Team conventions text. Sources: runner:crews.
  • Enforced policy [unsupported] No gate in the product (bounded search). Sources: runner:srch-pr.

Sources: runner:crews, runner:srch-pr

Agent cross-checked partial

No Git operations of its own; shows worktrees, status and diffs and lets agents run Git per the goal.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • branches: Shows branch and commits ahead.
  • worktrees: Picker lists existing worktrees; none created.
  • commits: Agents commit if authorized by goal.
  • pull_requests: None.
  • comments: None.
  • ci_response: None.
  • rebase: None.
  • conflicts: Conflict badges in Changes.
  • product_state: Git state read on demand.
9 atomic answers

Sources: runner:chats, runner:missions, runner:build-crew, runner:srch-pr

Agent cross-checked unsupported

No approve, merge, release or deploy authority.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • approve: None (bounded search).
  • merge: None (bounded search).
  • release: None (bounded search).
  • deploy: None (bounded search).
  • exact_revision: None (bounded search).
  • authority_enforcement: None (bounded search).
6 atomic answers
  • Approve [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Merge [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Release [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Deploy [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Exact revision [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Authority enforcement [unsupported] None (bounded search). Sources: runner:srch-pr.

Sources: runner:srch-pr

Agent cross-checked unsupported

No delivery evidence or rollback.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • commit_links: None (bounded search).
  • pr_links: None (bounded search).
  • build_links: None (bounded search).
  • deployment_links: None (bounded search).
  • delivery_status: None (bounded search).
  • retained_evidence: None (bounded search).
  • rollback: None (bounded search).
  • revert: None (bounded search).
  • stale_approval_invalidation: None (bounded search).
9 atomic answers
  • Delivery status [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Retained evidence [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Rollback [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Revert [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Stale approval invalidation [unsupported] None (bounded search). Sources: runner:srch-pr.

Sources: runner:srch-pr

Agent cross-checked partial

No batch review; per-file mark-viewed and a read-only diff.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • review_queue: None (bounded search).
  • summaries: None (bounded search).
  • grouping: None (bounded search).
  • risk_size_filters: None (bounded search).
  • findings: None (bounded search).
  • artifacts: None (bounded search).
  • request_changes: None (bounded search).
  • exact_revision_approval: None (bounded search).
  • dependency_queues: None (bounded search).
  • next_actions: Mark viewed advances to the next file.
  • stale_approval_invalidation: None (bounded search).
  • check_deduplication: None (bounded search).
12 atomic answers
  • Review queue [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Summaries [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Grouping [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Risk size filters [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Findings [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Artifacts [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Request changes [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Exact revision approval [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Dependency queues [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Next actions [partial] Mark viewed advances to the next file. Sources: runner:chats.
  • Stale approval invalidation [unsupported] None (bounded search). Sources: runner:srch-pr.
  • Check deduplication [unsupported] None (bounded search). Sources: runner:srch-pr.

Sources: runner:chats, runner:srch-pr

Ownership, privacy and economics

Agent cross-checked partial

MIT for the whole repository (desktop app, daemon, CLI); the relay service is closed and hosted; versions from 2026-08-22 to 2026-09-22 were GPL-3.0-only.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • server_license: Daemon and CLI are MIT; relay server is not in the public repository.
  • hosted_extensions: Hosted relay and account service closed.
  • hosted_parity: Hosted relay not inspectable.
6 atomic answers
  • Client license [supported] MIT. Sources: runner:license.
  • Server license [partial] Daemon and CLI are MIT; relay server is not in the public repository. Sources: runner:license, runner:remote.
  • Hosted extensions [partial] Hosted relay and account service closed. Sources: runner:remote.
  • Source visibility [supported] Public repository. Sources: runner:repoapi.
  • Source revision [supported] v0.20.0 commit c6ce633. Sources: runner:release.
  • Hosted parity [unknown] Hosted relay not inspectable. Sources: runner:remote.

Sources: runner:license, runner:repoapi, runner:remote, runner:release

Agent cross-checked partial

Fully local data (SQLite and event logs); works without any account; relay optional; no export feature documented.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • vendor_dependencies: Optional relay/account; Sparkle/appcast updates from releases.
  • export: None.
  • import: None.
  • formats: SQLite and event-log files; agent conversations in agent stores.
  • cancellation: Relay account terms not published.
  • history_portability: Agent histories live in agent folders.
8 atomic answers
  • Self hosting [supported] Runs locally with no vendor service. Sources: runner:arch, runner:remote.
  • Vendor dependencies [partial] Optional relay/account; Sparkle/appcast updates from releases. Sources: runner:remote, runner:install.
  • Export [unsupported] None. Sources: runner:srch-issue.
  • Import [unsupported] None. Sources: runner:srch-issue.
  • Formats [partial] SQLite and event-log files; agent conversations in agent stores. Sources: runner:arch.
  • Cancellation [unknown] Relay account terms not published.
  • Account independence [supported] Everything but remote access works signed out. Sources: runner:remote.
  • History portability [partial] Agent histories live in agent folders. Sources: runner:arch.

Sources: runner:arch, runner:remote, runner:install, runner:srch-issue

Agent cross-checked partial

Sessions and history stay on the machine; remote relay carries end-to-end encrypted traffic; no telemetry code found; no privacy policy page.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • credentials: Agent logins stay in agent configs; relay keys stored in OS keystore on release builds; PIN not sent to the service.
  • retention: No retention terms for relay metadata.
  • deletion: Delete missions; Reset remote access clears identity.
  • training: Provider policies third-party.
  • residency: Relay region not stated beyond mainland China.
  • subprocessors: No list.
  • telemetry_disclosure: No privacy page (404); docs silent.
  • telemetry_controls: None because no telemetry exists.
  • consent_prompt: No first-run consent dialog found.
15 atomic answers
  • Code data [supported] Local; goes only to the agents providers. Sources: runner:readme.
  • Prompts [supported] Local. Sources: runner:readme.
  • Credentials [partial] Agent logins stay in agent configs; relay keys stored in OS keystore on release builds; PIN not sent to the service. Sources: runner:remote.
  • Logs [supported] Local log folder. Sources: runner:arch.
  • Artifacts [supported] Local. Sources: runner:readme.
  • Retention [unknown] No retention terms for relay metadata.
  • Deletion [partial] Delete missions; Reset remote access clears identity. Sources: runner:remote.
  • Training [unknown] Provider policies third-party.
  • Residency [unknown] Relay region not stated beyond mainland China. Sources: runner:remote.
  • Subprocessors [unknown] No list.
  • Telemetry disclosure [unknown] No privacy page (404); docs silent. Sources: runner:site.
  • Telemetry defaults [supported] No telemetry code in the pinned release. Sources: runner:srch-tele.
  • Telemetry controls [unsupported] None because no telemetry exists. Sources: runner:srch-tele.
  • Provider separation [supported] Agent providers handled by agent CLIs. Sources: runner:install.

Sources: runner:readme, runner:remote, runner:srch-tele, runner:arch, runner:site, runner:install

Agent cross-checked partial

Free MIT software; donations via Ko-fi and Afdian; no paid plan found; remote access service pricing unpublished.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • currency: No price.
  • interval: No subscription.
  • seats: No seats.
  • seat_minimum: None.
  • credits: None.
  • trial: None; free.
  • overages: None.
  • taxes: Not applicable.
12 atomic answers

Sources: runner:readme, runner:license, runner:site, runner:install

Agent cross-checked partial

The app costs nothing; scenario costs are the agent subscriptions or API usage of each developer.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • one_person: App cost zero; agent costs separate.
  • two_people: Two independent installs; no collaboration features.
  • five_people: Five independent installs.
  • task_assumptions: No scenario assumptions.
  • concurrency: None.
  • tokens: None.
  • compute_hours: User machines.
  • variable_cost: Agent costs.
  • included_cost: None.
  • unknown_cost: Relay service pricing unknown.
11 atomic answers
  • One person [partial] App cost zero; agent costs separate. Sources: runner:readme, runner:install.
  • Two people [partial] Two independent installs; no collaboration features. Sources: runner:readme, runner:srch-team.
  • Five people [partial] Five independent installs. Sources: runner:readme.
  • Task assumptions [unknown] No scenario assumptions.
  • Concurrency [unknown] None.
  • Tokens [unknown] None.
  • Compute hours [unknown] User machines.
  • Fixed cost [supported] Zero for the app. Sources: runner:readme.
  • Variable cost [partial] Agent costs. Sources: runner:install.
  • Included cost [unsupported] None. Sources: runner:readme.
  • Unknown cost [partial] Relay service pricing unknown. Sources: runner:remote.

Sources: runner:readme, runner:install, runner:srch-team, runner:remote

Agent cross-checked partial

Alpha with releases every few days, public issues and Discord; no SLA or status page.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • response_commitments: None stated.
  • status_history: No status page.
  • deprecation: License changed twice (GPL window 2026-08-22 to 2026-09-22).
  • export_options: None.
8 atomic answers
  • Release stage [supported] README labels the product Alpha; Linux beta; remote access and iPhone private beta. Sources: runner:readme, runner:site.
  • Access restrictions [supported] Remote access invite-only, mainland China. Sources: runner:remote.
  • Support [supported] Discord, GitHub issues, WeChat group. Sources: runner:readme.
  • Response commitments [unsupported] None stated. Sources: runner:readme.
  • Release activity [supported] v0.19.8 2026-11-12, v0.19.9 2026-11-14, v0.20.0 2026-11-16 plus nightlies. Sources: runner:release.
  • Status history [unknown] No status page.
  • Deprecation [partial] License changed twice (GPL window 2026-08-22 to 2026-09-22). Sources: runner:license.
  • Export options [unsupported] None. Sources: runner:srch-issue.

Sources: runner:release, runner:repoapi, runner:readme, runner:site, runner:remote, runner:license, runner:srch-issue

No sourced public finding not-published

No publishable sourced finding is available for this criterion. This does not mean the product lacks the capability.

Agent cross-checked partial

Free open-source download without caps; remote access is invite-only beta.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • trial: No trial; free.
  • credits_trial: None.
  • paid_only: Nothing paid-only found.
  • free_scenarios: Local parallel agents and review free; no collaboration, issue management or remote compute.
  • card_required: None.
  • usage_caps: None.
  • paid_evaluation: None.
10 atomic answers

Sources: runner:readme, runner:remote, runner:install, runner:site, runner:srch-team

Agent cross-checked partial

No ADE price exists; remote access pricing unpublished.

Scope
Surfaces: Runner desktop app (Rust/GPUI) for macOS, Windows and Linux beta, runner CLI over the runnerd background service · Version: v0.20.0 (commit c6ce633, released 2026-11-16; alpha) · Plan: Open source (MIT); free · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-16 (6 days old)
Agent cross-check
2026-11-17
  • units: None.
  • included_limits: None.
  • overages: None.
  • compute_markup: None.
  • model_markup: None.
  • minimum_commitment: None.
  • cancellation: No subscription.
  • sales_only_unknowns: Relay service terms unpublished.
  • transparency_classification: Free application; relay terms not public.
9 atomic answers

Sources: runner:readme, runner:remote

Sources

runner:readme

Runner README

Native macOS/Windows (Linux beta) app where Claude Code, Codex, Antigravity, Copilot CLI, pi and others run in real terminals; reusable crews and missions with a shared feed; runner CLI; remote control private beta; MIT.

Kind
documented
Locator
Features; Also included; Supported agents; License
Source date
Not stated
Retrieved
2026-11-16
runner:install

Install Runner

macOS 12+, Windows 10 1809+ x64, Linux beta (glibc 2.35+, ARM64 and x86_64); agent CLIs provide their own model connection and accounts; first-run Setup detects agents.

Kind
documented
Locator
Download the app table; Set up an agent
Source date
Not stated
Retrieved
2026-11-16
runner:cli-doc

CLI reference

runner commands: remote, daemon, status, project, role, crew, mission (start, stop, resume, feed, answer, link), chat start, session (list, stop, resume, restart), msg; --json output; works with app closed.

Kind
documented
Locator
Command overview
Source date
Not stated
Retrieved
2026-11-16
runner:drive

Drive Runner from agents

Installs a Runner skill for detected agents so an agent can start missions, follow the feed and message members; missions started from a chat link back and report.

Kind
documented
Locator
Discover the installed version; Start and follow a crew
Source date
Not stated
Retrieved
2026-11-16
runner:missions

Missions

Mission = goal, per-member terminals, persistent feed; states Needs you/Failed/Interrupted/Working/Settled/Ended; shared directory, no branches or worktrees; mission launches use broad-access permissions; stop, resume, restart slot, archive.

Kind
documented
Locator
Goal and membership; Feed; Linked chat and reports; Permissions; Stop, resume and archive
Source date
Not stated
Retrieved
2026-11-16
runner:chats

Chats and windows

Side panel Files with git status and worktree picker, read-only Changes view with per-file diffs and mark-viewed, desktop notifications, split tabs and windows.

Kind
documented
Locator
Files view; Changes; Notifications
Source date
Not stated
Retrieved
2026-11-16
runner:remote

Remote access

Private beta, invite-only, mainland China; hosted relay with end-to-end encrypted sessions and PIN pairing; signed-out Runner works locally only; remote access never stops sessions.

Kind
documented
Locator
Sign in; Through Runner relay; PIN pairing
Source date
Not stated
Retrieved
2026-11-16
runner:site

runnersh.dev home page

Site banner: Runner for iPhone and remote access are in private beta with an iPhone waitlist; links to Docs and Release notes; no pricing page (404).

Kind
documented
Locator
banner
Source date
Not stated
Retrieved
2026-11-16
runner:release

Runner v0.20.0 release

Stable release 2026-11-16 with signed macOS arm64/x64 DMGs, Windows x64 setup exe, Linux arm64/x64 deb and tarballs, appcast feeds; v0.19.9 on 2026-11-14, v0.19.8 on 2026-11-12; a rolling nightly prerelease.

Kind
documented
Locator
release assets
Source date
2026-11-16
Retrieved
2026-11-16
runner:srch-team

Bounded search: organization|rbac|sso|oidc|saml|invite member|presence|co-edit

No multi-user collaboration, roles or SSO; account sign-in exists only for the relay and pairing is per-account computers.

Kind
source-inspected
Locator
grep -rniE "organization|rbac|sso|oidc|saml|invite member|presence|co-edit" over crates docs/guide README.md at v0.20.0
Source date
Not stated
Retrieved
2026-11-16
Commit
c6ce6337e6c49a67dd4432b414ed739e1ef55ab0
Path
crates
runner:agents-doc

Supported agents

Codex, Claude Code, Antigravity CLI, pi, Copilot CLI, Cursor, Grok Build, OpenCode v2 and opt-in TRAE; separately installed CLIs keep native TUI; per-agent capability table; Bypass defaults off for chats; missions use broad access.

Kind
documented
Locator
Install and sign in; Capabilities; Launch options
Source date
Not stated
Retrieved
2026-11-16
runner:crews

Crews

Reusable teams of role slots with unique handles, per-slot agent/model/effort overrides and shared team conventions; starter Pair coding crew.

Kind
documented
Locator
Slots; Team conventions
Source date
Not stated
Retrieved
2026-11-16
runner:arch

Architecture

Three-process model: GPUI app, runnerd daemon (SQLite, session manager, event bus) and runner CLI over runnerd.sock; sessions survive app quit; usage module reads Claude Code, Codex and Antigravity limits; Claude/Codex update checks disabled inside sessions.

Kind
source-inspected
Locator
Overview; usage module; agent updates; mission inbox
Source date
Not stated
Retrieved
2026-11-16
Commit
c6ce6337e6c49a67dd4432b414ed739e1ef55ab0
Path
docs/arch/arch.md
runner:claude-acc

Claude Code accounts

Multiple Claude Code sign-ins in separate config folders with a usage popover showing weekly usage.

Kind
documented
Locator
Add an account; Switch the active account
Source date
Not stated
Retrieved
2026-11-16
runner:srch-cost

Bounded search: cost|dollar|price per|billing

No per-task or dollar cost accounting; usage shows plan-limit percentages for Claude Code, Codex and Antigravity CLI.

Kind
source-inspected
Locator
grep -rniE "cost|dollar|price per|billing" over crates docs/guide README.md at v0.20.0
Source date
Not stated
Retrieved
2026-11-16
Commit
c6ce6337e6c49a67dd4432b414ed739e1ef55ab0
Path
crates
runner:first-mission

First mission

Runner starts the crew in the directory given and does not create an isolated copy or Git worktree.

Kind
documented
Locator
directory note
Source date
Not stated
Retrieved
2026-11-16
runner:srch-pr

Bounded search: pull request|gh pr|merge request|git push|git commit

No PR, push or commit operations in the product; the mission goal tells agents which Git actions are authorized.

Kind
source-inspected
Locator
grep -rniE "pull request|gh pr|merge request|git push|git commit" over crates docs/guide README.md at v0.20.0
Source date
Not stated
Retrieved
2026-11-16
Commit
c6ce6337e6c49a67dd4432b414ed739e1ef55ab0
Path
crates
runner:process

Process model

Daemon keeps running without clients; Ask/Keep/Stop on app quit; app crash leaves daemon sessions running; daemon crash closes PTYs on macOS and kills sessions.

Kind
source-inspected
Locator
runnerd lifecycle
Source date
Not stated
Retrieved
2026-11-16
Commit
c6ce6337e6c49a67dd4432b414ed739e1ef55ab0
Path
docs/arch/process-model.md
runner:srch-sched

Bounded search: cron|schedule|recurring|trigger

No scheduler, cron or event-trigger feature; matches are internal refresh timers (usage polling) and hook triggers.

Kind
source-inspected
Locator
grep -rniE "cron|schedule|recurring|trigger" over crates docs/guide README.md at v0.20.0
Source date
Not stated
Retrieved
2026-11-16
Commit
c6ce6337e6c49a67dd4432b414ed739e1ef55ab0
Path
crates
runner:mcp-skills

MCP servers and skills

Settings pages list and toggle each agent own MCP servers and skills, editing the native config for new sessions.

Kind
documented
Locator
MCP servers; Skills
Source date
Not stated
Retrieved
2026-11-16
runner:cargo

Workspace Cargo.toml

Rust workspace version 0.20.0 with crates runner-app, runner-cli, runner-core, runner-daemon, runner-terminal; vendored dependencies under vendor/.

Kind
source-inspected
Locator
workspace package version
Source date
Not stated
Retrieved
2026-11-16
Commit
c6ce6337e6c49a67dd4432b414ed739e1ef55ab0
Path
Cargo.toml
runner:srch-issue

Bounded search: issue tracker|ticket|backlog|kanban|epic|todo list

No issue, ticket or backlog entity; missions are sessions with a goal, not trackable work items.

Kind
source-inspected
Locator
grep -rniE "issue tracker|ticket|backlog|kanban|epic|todo list" over crates docs/guide README.md at v0.20.0
Source date
Not stated
Retrieved
2026-11-16
Commit
c6ce6337e6c49a67dd4432b414ed739e1ef55ab0
Path
crates
runner:sessions

Sessions

A session is one agent or shell in a real terminal; hooks or terminal activity report state; stop, resume, restart actions.

Kind
documented
Locator
Read the state; Choose the right action
Source date
Not stated
Retrieved
2026-11-16
runner:srch-limits

Bounded search: max_turns|token limit|spend limit|budget|timeout

No turn, token or spend limit enforcement for agents; only operational timeouts for Runner internals (30 s sign-in, 10-minute draft hold).

Kind
source-inspected
Locator
grep -rniE "max_turns|token limit|spend limit|budget|timeout" over crates docs/guide README.md at v0.20.0
Source date
Not stated
Retrieved
2026-11-16
Commit
c6ce6337e6c49a67dd4432b414ed739e1ef55ab0
Path
crates
runner:build-crew

Build a crew

Runner does not isolate working directories automatically; create a Git worktree before launch if needed.

Kind
documented
Locator
working directories note
Source date
Not stated
Retrieved
2026-11-16
runner:linux-service

Runner Linux packages and optional runnerd user service

Linux packages ship runner, runnerd and runner-app plus an optional systemd user service for runnerd; troubleshooting says that if the CLI is used over SSH, start runnerd with systemctl --user first. This documents running the daemon and CLI on a Linux host reached over SSH, but no guide presents remote or server execution as a supported workflow.

Kind
source-inspected
Locator
Optional systemd user service (lines 92-107); docs/guide/guides/troubleshooting.md Linux local builds (line 76); packaging/linux/runnerd.service
Source date
Not stated
Retrieved
2026-11-17
Commit
c6ce6337e6c49a67dd4432b414ed739e1ef55ab0
Path
docs/arch/linux.md
runner:srch-tele

Bounded search: telemetry|analytics|sentry|posthog|mixpanel|amplitude|plausible

No telemetry or analytics code in the Rust crates or docs; the app makes update, usage, relay and agent-CLI calls only.

Kind
source-inspected
Locator
grep -rniE "telemetry|analytics|sentry|posthog|mixpanel|amplitude|plausible" over crates docs/guide README.md at v0.20.0
Source date
Not stated
Retrieved
2026-11-16
Commit
c6ce6337e6c49a67dd4432b414ed739e1ef55ab0
Path
crates
runner:license

Runner LICENSE

MIT, copyright 2026 Yicheng Wang; README notes versions released 2026-08-22 to 2026-09-22 were GPL-3.0-only.

Kind
documented
Locator
MIT License text
Source date
Not stated
Retrieved
2026-11-16
runner:repoapi

GitHub repository API

521 stars on 2026-11-16; MIT license detected; pushed 2026-11-16; homepage https://runnersh.dev/.

Kind
documented
Locator
stargazers_count, license, pushed_at
Source date
Not stated
Retrieved
2026-11-16

Comparisons with Runner

Compare Runner with any other product on adecomparison.com.