50 of 51 criteria have a sourced finding; 50 were also cross-checked by a second agent. Every comparison that includes LoopTroop 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.
Browser UI served by a local daemon at 127.0.0.1:3000 (macOS, Linux, Windows, WSL), plus a `looptroop` CLI for daemon control; no desktop or phone app; remote browser access is an opt-in HTTPS reverse-proxy setup.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace), looptroop CLI (daemon control, doctor, setup, clean) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
desktop_os: No native desktop app; the daemon and browser UI run on macOS (arm64 executable, Intel via npm or Homebrew), Linux x64/arm64 and Windows x64 (Windows ARM via npm or Scoop).
mobile_apps: No iOS or Android app.
mobile_browser: UI has narrow-screen layouts (search popover, wrapping controls) but the daemon binds loopback by default; phone access needs the documented HTTPS public-origin proxy setup.
cli: `looptroop` CLI starts, stops, opens, diagnoses and cleans the daemon; ticket work is not driven from the CLI.
editor_extensions: No editor extension.
notifications: Board shows Needs Input and error badges and a blue pulse for AI questions; no browser, push or chat notifications.
cross_device_continuity: State lives in the daemon and SQLite so another browser can reconnect to the same ticket; no accounts or sync, and remote access needs a self-configured proxy.
9 atomic answers
Desktop os [partial] No native desktop app; the daemon and browser UI run on macOS (arm64 executable, Intel via npm or Homebrew), Linux x64/arm64 and Windows x64 (Windows ARM via npm or Scoop). Sources: looptroop:install.
Browser [supported] Full UI (Kanban board, ticket workspace, approvals, diffs, logs) is a browser app served by the daemon. Sources: looptroop:frontend, looptroop:cli.
Mobile browser [partial] UI has narrow-screen layouts (search popover, wrapping controls) but the daemon binds loopback by default; phone access needs the documented HTTPS public-origin proxy setup. Sources: looptroop:frontend, looptroop:config.
Cli [partial] `looptroop` CLI starts, stops, opens, diagnoses and cleans the daemon; ticket work is not driven from the CLI. Sources: looptroop:clisrc, looptroop:cli.
Notifications [partial] Board shows Needs Input and error badges and a blue pulse for AI questions; no browser, push or chat notifications. Sources: looptroop:ticketflow, looptroop:srch-trigger.
Cross device continuity [partial] State lives in the daemon and SQLite so another browser can reconnect to the same ticket; no accounts or sync, and remote access needs a self-configured proxy. Sources: looptroop:ticketflow, looptroop:config.
Single embedded harness: OpenCode (v1 or v2 server, supervised or adopted). Any provider and model OpenCode exposes can fill the implementer and council slots; no other coding harness.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
harnesses: OpenCode is the only supported harness; no Claude Code, Codex, Gemini CLI or other adapter.
custom_adapters: No adapter API for other harnesses; only the OpenCode transports exist.
mixed_harness_operation: One harness only; mixed-model councils exist but all run through OpenCode.
5 atomic answers
Harnesses [partial] OpenCode is the only supported harness; no Claude Code, Codex, Gemini CLI or other adapter. Sources: looptroop:oc, looptroop:srch-harness.
Model providers [supported] Providers and models come from the user OpenCode installation (OpenAI, Anthropic, Google, OpenRouter, NVIDIA NIM and others) and are selected per slot with effort variants. Sources: looptroop:config, looptroop:start.
Custom adapters [unsupported] No adapter API for other harnesses; only the OpenCode transports exist. Sources: looptroop:srch-harness.
Session integration [supported] Controls OpenCode through its server API (sessions, prompts, SSE events, permissions) and starts `opencode serve` when none is reachable. Sources: looptroop:oc.
Mixed harness operation [unsupported] One harness only; mixed-model councils exist but all run through OpenCode. Sources: looptroop:srch-harness, looptroop:llm.
No LoopTroop account or credits; model access is whatever OpenCode is configured with: API keys, OpenRouter/NVIDIA free tiers, OpenCode Go, or a Codex subscription through OpenCode.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
agent_subscriptions: Docs name a Codex subscription through OpenCode; availability depends on OpenCode login support and is not implemented by LoopTroop.
vendor_credits: LoopTroop sells no credits or usage.
local_models: Local models are not documented by LoopTroop; any OpenCode-supported local provider would be inherited, but no LoopTroop page names one.
account_switching: No multi-account switching feature in LoopTroop; behaviour is OpenCode provider configuration.
provider_restrictions: Provider terms on automated unattended use are vendor policy facts outside the repository.
7 atomic answers
Agent subscriptions [partial] Docs name a Codex subscription through OpenCode; availability depends on OpenCode login support and is not implemented by LoopTroop. Sources: looptroop:start, looptroop:oc.
Api accounts [supported] API keys for OpenAI, Anthropic, Google, OpenRouter and NVIDIA NIM configured in OpenCode. Sources: looptroop:start.
Vendor credits [unsupported] LoopTroop sells no credits or usage. Sources: looptroop:website.
Local models [unknown] Local models are not documented by LoopTroop; any OpenCode-supported local provider would be inherited, but no LoopTroop page names one. Sources: looptroop:oc.
Account switching [unknown] No multi-account switching feature in LoopTroop; behaviour is OpenCode provider configuration. Sources: looptroop:oc.
Vendor account requirement [supported] No LoopTroop account needed; a GitHub login (`gh`) and an OpenCode provider credential are required. Sources: looptroop:start.
Provider restrictions [unknown] Provider terms on automated unattended use are vendor policy facts outside the repository. Sources: looptroop:start.
Install via curl/PowerShell, npm, Homebrew, Scoop, Chocolatey, WinGet, Docker or executable; then `looptroop open`, configure OpenCode, pick models, attach a local repo with a GitHub origin and create a ticket. A GitHub origin and `gh` are mandatory.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace), looptroop CLI (daemon control, doctor, setup, clean) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
no_git: A project must be a valid git repository with a GitHub origin.
local_git_without_remote: Attach verifies a GitHub origin and pre-flight checks `gh`; a repo without a remote cannot finish the PR flow.
managed_cloud_setup: No managed cloud exists.
byo_vps_onboarding: README and docs recommend a disposable VM or cloud dev machine and document remote-access environment variables and a Docker image, but there is no VPS installer or onboarding flow.
7 atomic answers
First task setup [supported] Install, `looptroop open`, configure OpenCode provider, choose models in Configuration, Add Project path, create ticket; `looptroop setup` can attach a project from the terminal. Sources: looptroop:start, looptroop:clisrc.
No git [unsupported] A project must be a valid git repository with a GitHub origin. Sources: looptroop:start.
Local git without remote [unsupported] Attach verifies a GitHub origin and pre-flight checks `gh`; a repo without a remote cannot finish the PR flow. Sources: looptroop:start, looptroop:post.
Hosted git [supported] GitHub-backed repos are the supported route (push and draft PR through `gh`); other hosts are not documented. Sources: looptroop:start, looptroop:post.
Managed cloud setup [unsupported] No managed cloud exists. Sources: looptroop:srch-cloud.
Byo vps onboarding [partial] README and docs recommend a disposable VM or cloud dev machine and document remote-access environment variables and a Docker image, but there is no VPS installer or onboarding flow. Sources: looptroop:readme, looptroop:config, looptroop:install.
Local REST API (same one the UI uses) with bearer-token auth plus SSE; CLI only controls the daemon. No SDK, MCP server or webhooks.
Scope
Surfaces: Local REST API and SSE stream under /api, looptroop CLI (daemon control, doctor, setup, clean) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
cli: CLI covers open/start/stop/status/logs/doctor/setup/clean; ticket create/steer/review is not in the CLI.
sdk: No published SDK.
mcp_server: LoopTroop does not expose an MCP server.
webhooks: No webhook support.
stability: Docs call it the current HTTP surface; the project is early alpha and 0.6.4 broke compatibility, with no stability promise.
8 atomic answers
Cli [partial] CLI covers open/start/stop/status/logs/doctor/setup/clean; ticket create/steer/review is not in the CLI. Sources: looptroop:clisrc, looptroop:cli.
Api [supported] REST routes under /api for projects, tickets, workflow actions, artifacts, beads, files, logs and prompts; /api/stream SSE. Sources: looptroop:api.
Sdk [unsupported] No published SDK. Sources: looptroop:pkg.
Mcp server [unsupported] LoopTroop does not expose an MCP server. Sources: looptroop:srch-mcp.
Operations [supported] API can create tickets, start, approve, retry/continue, cancel, merge, answer questions, read artifacts and logs and edit plans. Sources: looptroop:api, looptroop:ticketflow.
Authentication [supported] Bearer token minted at startup in owner-readable daemon.json, or HttpOnly session cookie; rate limits per client. Sources: looptroop:api.
Stability [partial] Docs call it the current HTTP surface; the project is early alpha and 0.6.4 broke compatibility, with no stability promise. Sources: looptroop:api, looptroop:changelog.
MCP servers, skills and provider configuration are inherited from the user OpenCode setup; LoopTroop itself has no plugin or custom-tool system but lets users edit prompts and global rules.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
mcp_client: Inherited: OpenCode MCP configuration remains available to LoopTroop sessions; LoopTroop adds no MCP management.
plugins: No LoopTroop plugin system.
custom_tools: Tool access is whatever OpenCode provides, constrained by LoopTroop tool-policy and permission guard.
instructions_import: Editable prompt templates and global rule blocks via UI and /api/prompts; repository AGENTS.md handling is left to OpenCode.
authentication: Integration credentials are held in OpenCode; GitHub uses `gh` auth and GH_TOKEN passed to child processes.
permission_scope: Execution sessions run allow-all; a guard refuses direct GitHub writes, remote git writes and destructive changes to managed worktree/ticket state.
7 atomic answers
Mcp client [partial] Inherited: OpenCode MCP configuration remains available to LoopTroop sessions; LoopTroop adds no MCP management. Sources: looptroop:oc.
Plugins [unsupported] No LoopTroop plugin system. Sources: looptroop:srch-mcp.
Custom tools [partial] Tool access is whatever OpenCode provides, constrained by LoopTroop tool-policy and permission guard. Sources: looptroop:oc.
Instructions import [partial] Editable prompt templates and global rule blocks via UI and /api/prompts; repository AGENTS.md handling is left to OpenCode. Sources: looptroop:api, looptroop:oc.
Authentication [partial] Integration credentials are held in OpenCode; GitHub uses `gh` auth and GH_TOKEN passed to child processes. Sources: looptroop:ops.
Permission scope [partial] Execution sessions run allow-all; a guard refuses direct GitHub writes, remote git writes and destructive changes to managed worktree/ticket state. Sources: looptroop:permissions, looptroop:ops.
Integration owner [supported] OpenCode provides MCP/skills/providers; LoopTroop provides prompts and the git/GitHub integration. Sources: looptroop:oc.
Web UI built with React and Vite, served by a Node daemon (TypeScript, Hono, SQLite); no desktop or mobile shell.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Tickets are created inside LoopTroop (title, description, priority) and are the authoritative record; no import from GitHub, Linear or Jira and no chat or Git triggers.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
external_import: No tracker import.
chat_git_triggers: Work starts only by a manual Start; no chat or Git event triggers.
sync: No external sync; the GitHub PR is the only external object.
external_roundtrips: GitHub is needed only to merge the PR (or merge manually); no tracker round trips.
9 atomic answers
Local creation [supported] New Ticket form creates a DRAFT backlog ticket with a composite ref such as 1:AUTH-12. Sources: looptroop:ticketflow, looptroop:api.
Prompt intake [supported] Ticket description is the prompt; an AI interview then clarifies it. Sources: looptroop:readme.
Chat git triggers [unsupported] Work starts only by a manual Start; no chat or Git event triggers. Sources: looptroop:srch-trigger.
Authoritative tracker [supported] LoopTroop project SQLite and .ticket artifacts are authoritative. Sources: looptroop:ops.
Stable ids [supported] Public composite ticket refs projectId:externalId are stable and used by routes. Sources: looptroop:api.
Sync [unsupported] No external sync; the GitHub PR is the only external object. Sources: looptroop:srch-import.
Supported edits [supported] Draft, interview, PRD, beads and setup-plan artifacts are editable before PRE_FLIGHT_CHECK with hash-guarded saves. Sources: looptroop:ticketflow, looptroop:beads.
External roundtrips [partial] GitHub is needed only to merge the PR (or merge manually); no tracker round trips. Sources: looptroop:post.
Durable per-ticket plan chain: interview.yaml, prd.yaml (epics, user stories, steps) and beads JSONL with acceptance criteria, all editable at approval gates with archived phase attempts and edit receipts.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
amendment: Planning edits are accepted only before PRE_FLIGHT_CHECK; setup-plan edits rewind during preparation; running coding is not plan-amendable.
Acceptance criteria [supported] Each bead has measurable acceptance criteria and expected tests. Sources: looptroop:beads, looptroop:readme.
Editors [supported] Structured and raw editors for interview, PRD, beads and setup plan with autosave and hash-guarded save. Sources: looptroop:beads, looptroop:frontend.
Implementation links [supported] Beads map to bead commits and diffs (`bead_diff:{id}`, bead commits) and a squashed candidate. Sources: looptroop:beads, looptroop:post.
Amendment [partial] Planning edits are accepted only before PRE_FLIGHT_CHECK; setup-plan edits rewind during preparation; running coding is not plan-amendable. Sources: looptroop:ticketflow.
Restart persistence [supported] Artifacts persist in worktree and SQLite and the XState snapshot is rehydrated on restart. Sources: looptroop:ticketflow.
Tickets decompose into PRD epics and user stories and then dependency-ordered beads (`blocked_by`, priority) that the scheduler enforces; no arbitrary ticket hierarchy or growing a ticket into an epic.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
epics: PRD contains epics and user stories; tickets themselves are flat (Manual QA can spawn child tickets).
hierarchy: Ticket → epics/stories → beads; Manual QA improvements create new tickets; no arbitrary depth.
state_propagation: Bead status pending/in_progress/done/error drives ticket progress; error beads are not auto-selected.
task_to_epic: A ticket cannot grow into an epic after start; planning restarts rewrite downstream artifacts.
identity_retention: Planning edits keep the ticket ID and archive earlier generations but invalidate downstream artifacts.
workflow_constraints: Fixed pipeline for every ticket (interview, PRD, beads, setup, coding, tests, PR); Manual QA is optional.
9 atomic answers
Epics [partial] PRD contains epics and user stories; tickets themselves are flat (Manual QA can spawn child tickets). Sources: looptroop:readme, looptroop:ticketflow.
Hierarchy [partial] Ticket → epics/stories → beads; Manual QA improvements create new tickets; no arbitrary depth. Sources: looptroop:readme, looptroop:ticketflow.
Blockers [supported] dependencies.blocked_by with server-derived blocks, validated for cycles and dangling references. Sources: looptroop:beads.
Readiness [supported] getRunnable returns pending beads whose blockers are done; pre-flight requires at least one runnable bead. Sources: looptroop:scheduler, looptroop:beads.
Enforcement [supported] Scheduler never runs a bead before its blockers are done; the model does not pick order. Sources: looptroop:scheduler.
State propagation [partial] Bead status pending/in_progress/done/error drives ticket progress; error beads are not auto-selected. Sources: looptroop:beads, looptroop:scheduler.
Task to epic [unsupported] A ticket cannot grow into an epic after start; planning restarts rewrite downstream artifacts. Sources: looptroop:ticketflow.
Identity retention [partial] Planning edits keep the ticket ID and archive earlier generations but invalidate downstream artifacts. Sources: looptroop:ticketflow.
Workflow constraints [partial] Fixed pipeline for every ticket (interview, PRD, beads, setup, coding, tests, PR); Manual QA is optional. Sources: looptroop:ticketflow.
Per-ticket minimal context built from artifacts; editable prompt templates and global rules; OpenCode skills/AGENTS.md inherited. No cross-ticket memory or maintained project docs yet (roadmap only).
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
instructions: Global rule blocks and prompt templates are editable; repository instruction files (AGENTS.md) are read by OpenCode, not managed by LoopTroop.
skills: OpenCode skills remain available to sessions; LoopTroop adds none.
project_docs: Relevant-files scan and repo discovery artifacts per ticket; no maintained project doc store.
memory: Per-bead retry notes and Failed Iteration Notes persist within a ticket; cross-ticket memory is only on the roadmap.
scope: Context is scoped per status (active bead, file target, tests) by buildMinimalContext.
provenance: Artifacts carry content hashes and edit receipts; no citation system for context items.
correction: Users can edit planning artifacts and prompts; no memory correction UI.
deletion: Deleting a ticket or project cascades its artifacts and chat conversations.
freshness: Post-interview context refresh is roadmap-only; relevant-file scan runs at ticket start.
takeover_continuity: Artifacts and logs persist in the project directory for any user of that machine; no multi-user handoff.
10 atomic answers
Instructions [partial] Global rule blocks and prompt templates are editable; repository instruction files (AGENTS.md) are read by OpenCode, not managed by LoopTroop. Sources: looptroop:api, looptroop:oc.
Skills [partial] OpenCode skills remain available to sessions; LoopTroop adds none. Sources: looptroop:oc.
Project docs [partial] Relevant-files scan and repo discovery artifacts per ticket; no maintained project doc store. Sources: looptroop:ticketflow.
Memory [partial] Per-bead retry notes and Failed Iteration Notes persist within a ticket; cross-ticket memory is only on the roadmap. Sources: looptroop:beads, looptroop:roadmapdoc.
Scope [partial] Context is scoped per status (active bead, file target, tests) by buildMinimalContext. Sources: looptroop:readme.
Provenance [partial] Artifacts carry content hashes and edit receipts; no citation system for context items. Sources: looptroop:api.
Correction [partial] Users can edit planning artifacts and prompts; no memory correction UI. Sources: looptroop:beads, looptroop:api.
Deletion [partial] Deleting a ticket or project cascades its artifacts and chat conversations. Sources: looptroop:db.
Freshness [partial] Post-interview context refresh is roadmap-only; relevant-file scan runs at ticket start. Sources: looptroop:roadmapdoc.
Takeover continuity [partial] Artifacts and logs persist in the project directory for any user of that machine; no multi-user handoff. Sources: looptroop:ops.
Per-ticket relevant-files scan and codebase discovery artifacts, then focused context packs; no persistent index, external knowledge or citation enforcement.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
code_search: SCANNING_RELEVANT_FILES agent step searches the repo via OpenCode tools per ticket.
indexing: No persistent repository index in the pinned release (roadmap mentions future repo index).
external_knowledge: Only what OpenCode tools or MCP servers provide.
multimodal_inputs: Manual QA accepts evidence uploads; no image input to planning documented.
freshness: Scan happens at start; refresh after interview is roadmap-only.
api_retrieval: No API/documentation retrieval feature.
citations: Interview/PRD artifacts do not cite sources for context items.
conflict_detection: Council voting, coverage verification loops and structured checks detect gaps between ticket, PRD and beads.
review_enforcement: Coverage passes and a final candidate file audit check scope before the PR.
10 atomic answers
Code search [partial] SCANNING_RELEVANT_FILES agent step searches the repo via OpenCode tools per ticket. Sources: looptroop:ticketflow.
Indexing [unsupported] No persistent repository index in the pinned release (roadmap mentions future repo index). Sources: looptroop:roadmapdoc.
External knowledge [partial] Only what OpenCode tools or MCP servers provide. Sources: looptroop:oc.
Multimodal inputs [partial] Manual QA accepts evidence uploads; no image input to planning documented. Sources: looptroop:frontend.
Freshness [partial] Scan happens at start; refresh after interview is roadmap-only. Sources: looptroop:roadmapdoc.
Access scope [supported] Context is limited to the ticket worktree and the original checkout read-only. Sources: looptroop:ticketflow, looptroop:beads.
Api retrieval [unsupported] No API/documentation retrieval feature. Sources: looptroop:srch-mcp.
Citations [unsupported] Interview/PRD artifacts do not cite sources for context items. Sources: looptroop:ticketflow.
Conflict detection [partial] Council voting, coverage verification loops and structured checks detect gaps between ticket, PRD and beads. Sources: looptroop:llm, looptroop:ticketflow.
Review enforcement [partial] Coverage passes and a final candidate file audit check scope before the PR. Sources: looptroop:post, looptroop:ticketflow.
Kanban board of tickets with backlog (To Do), priority P1-P5, project, search and triage filters; ideas can sit as DRAFT until Start; one ticket spawns many sessions but there is no discussion or assignment.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
discuss: No comment threads; comments are roadmap-only.
label: No labels documented; filters use project, phase, status, priority, errors.
assign: No assignees; single user.
archive: Done and Canceled columns retain tickets; archive is not a separate feature.
reopen: Terminal COMPLETED/CANCELED tickets are not reopened; Manual QA spawns a new ticket.
project_epic_views: Board per all projects plus project filter; no epic view.
Agents do not operate the ticket tracker: tickets are driven by the daemon state machine; agents only edit worktree files and return structured outputs. The API and bead JSONL are shared by UI and scripts, not by agents.
Scope
Surfaces: Local REST API and SSE stream under /api · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
read: Prompts supply ticket/PRD/bead context to agents; agents have no tracker tool.
create: Agents do not create tickets (Manual QA improvements are created by the app).
update: Agents report completion through a structured BEAD_STATUS marker that the app records.
claim: No agent claim mechanism; scheduler assigns beads.
assign: Scheduler picks the next bead.
state_changes: Bead state changes are driven by the app from agent markers.
dependencies: Councils write bead dependencies during planning; no runtime dependency creation.
children: No runtime child creation by agents.
attribution: Logs attribute actions to models and sessions; no user attribution.
permissions: Agents have no tracker permissions model.
conflicts: Hash-guarded saves reject stale writes.
14 atomic answers
Shared ids [supported] Ticket and bead IDs are the same in UI and API. Sources: looptroop:api.
Read [partial] Prompts supply ticket/PRD/bead context to agents; agents have no tracker tool. Sources: looptroop:beads.
Create [unsupported] Agents do not create tickets (Manual QA improvements are created by the app). Sources: looptroop:srch-mcp.
Update [partial] Agents report completion through a structured BEAD_STATUS marker that the app records. Sources: looptroop:beads.
No living product specification: PRD is a per-ticket artifact; delta-merge living specs are roadmap-only.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
persistent_requirements: PRD is per ticket; living project specs are a roadmap item.
human_edits: PRD editable per ticket before pre-flight.
agent_edits: Council drafts and refines the per-ticket PRD.
revisions: Archived attempts and edit receipts for the ticket PRD.
implementation_links: Beads trace to PRD stories via coverage verification.
later_retrieval: Later tickets do not retrieve earlier PRDs.
drift_detection: No drift detection between spec and code across tasks.
review_integration: Within one ticket, final tests and audits reference the PRD.
9 atomic answers
Persistent requirements [unsupported] PRD is per ticket; living project specs are a roadmap item. Sources: looptroop:roadmapdoc.
Human edits [partial] PRD editable per ticket before pre-flight. Sources: looptroop:ticketflow.
Agent edits [partial] Council drafts and refines the per-ticket PRD. Sources: looptroop:llm.
Revisions [partial] Archived attempts and edit receipts for the ticket PRD. Sources: looptroop:ticketflow.
Council members and planning sessions run in parallel; only one ticket per project may be in the execution band (coding to cleanup); different projects and planning-stage tickets can run side by side.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
concurrent_tasks: Multiple tickets can plan at once; one execution-band ticket per project.
observed_concurrency: Measured concurrency is a benchmark fact; this catalog runs nothing.
5 atomic answers
Concurrent sessions [supported] Council drafting and voting run model sessions concurrently. Sources: looptroop:llm.
Concurrent tasks [partial] Multiple tickets can plan at once; one execution-band ticket per project. Sources: looptroop:ticketflow.
Observed concurrency [unknown] Measured concurrency is a benchmark fact; this catalog runs nothing. Sources: looptroop:ticketflow.
Status visibility [supported] Kanban columns To Do, In Progress, Needs Input, Done plus an Errors group and per-model chips. Sources: looptroop:ticketflow, looptroop:frontend.
Work is dispatched by the user pressing Start; the scheduler assigns beads deterministically; a project lock enforces exclusive execution. No assignment to humans.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
human_assignment: No human assignees.
agent_assignment: Main implementer and council models are chosen per ticket and locked at Start.
queues: A blocked second ticket waits at its gate and the user retries; no automatic queue.
routing: Fixed model slots per ticket; no routing rules.
7 atomic answers
Human assignment [unsupported] No human assignees. Sources: looptroop:srch-team.
Agent assignment [partial] Main implementer and council models are chosen per ticket and locked at Start. Sources: looptroop:config.
Claims [supported] Project execution lock prevents a second ticket entering execution. Sources: looptroop:ticketflow.
Queues [partial] A blocked second ticket waits at its gate and the user retries; no automatic queue. Sources: looptroop:ticketflow.
Routing [partial] Fixed model slots per ticket; no routing rules. Sources: looptroop:config.
Duplicate prevention [supported] One daemon per config dir and one execution ticket per project. Sources: looptroop:cli, looptroop:ticketflow.
Ownership enforcement [supported] Execution lock and session ownership records are enforced server-side. Sources: looptroop:ticketflow, looptroop:oc.
Product-owned multi-model orchestration: a deterministic state machine runs council drafters, voters and refiners, then coding sessions; there is no delegating coordinator agent or child-agent messaging.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
messages: Agents exchange no messages; the app passes artifacts between stages.
cross_harness: All models run through OpenCode.
authentication: Provider auth via OpenCode.
followup: Stop or retry one model during a step; no follow-up with a finished model.
12 atomic answers
Roles [supported] Main implementer, council members, voters, refiner, coverage verifier, setup agent and assistant have distinct roles. Sources: looptroop:llm, looptroop:config.
Model harness effort [supported] Per-slot model and effort variant; one harness. Sources: looptroop:config.
Messages [partial] Agents exchange no messages; the app passes artifacts between stages. Sources: looptroop:llm.
Results return [supported] Drafts, votes, refined plans and bead results are stored as artifacts. Sources: looptroop:llm, looptroop:ticketflow.
Cross harness [unsupported] All models run through OpenCode. Sources: looptroop:srch-harness.
Orchestration owner [supported] LoopTroop state machine, not an agent. Sources: looptroop:ticketflow.
Activation [supported] Councils are on by default for planning. Sources: looptroop:llm.
Authentication [partial] Provider auth via OpenCode. Sources: looptroop:oc.
Live progress [supported] Council chips, voting details, logs and SSE show progress live. Sources: looptroop:llm, looptroop:frontend.
Retained results [supported] Per-model drafts, scores and votes are retained as phase artifacts. Sources: looptroop:llm.
Followup [partial] Stop or retry one model during a step; no follow-up with a finished model. Sources: looptroop:llm.
Invocation [supported] Automatic as part of the ticket flow. Sources: looptroop:ticketflow.
No unattended triggers: tickets start manually; once started the pipeline runs unattended until a gate or error.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
schedules: No scheduler for starting tickets.
events: No event or webhook triggers.
recurring_tasks: No recurring tasks.
completion_conditions: Pipeline completes when final tests, optional QA and PR merge finish.
scheduler_location: No trigger scheduler exists.
overlaps: Not applicable without triggers.
8 atomic answers
Schedules [unsupported] No scheduler for starting tickets. Sources: looptroop:srch-trigger.
Strong gates and recovery: cancel at most statuses, stop or retry one council model, retry with note, continue preserved sessions, answer or skip AI questions; no mid-run steering of coding.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
pause: Human gates pause the flow; no generic pause button.
steer: Retry with extra note and answering AI questions; no free-form steering of a running coding session.
notifications: Board Needs Input status; no push channels.
escalation: BLOCKED_ERROR after retry budget; AI question wait timeout.
chat_terminal_parity: No terminal/chat steering surface; read-only assistant chat.
10 atomic answers
Pause [partial] Human gates pause the flow; no generic pause button. Sources: looptroop:ticketflow.
Stop [supported] Cancel at any cancelable status; stop one council model. Sources: looptroop:ticketflow.
Resume [supported] Retry or Continue from BLOCKED_ERROR and restart rehydration. Sources: looptroop:ticketflow.
Steer [partial] Retry with extra note and answering AI questions; no free-form steering of a running coding session. Sources: looptroop:ticketflow.
Kanban board across tickets and projects with progress rings and bead chips, an Errors group, filters, and a per-ticket phase navigator; no epic or dependency graph across tickets.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
parent_child: Bead lists within a ticket; Manual QA child tickets.
dependencies: Bead dependencies shown within a ticket only.
workload: Progress ring and bead counts; ETA computed from history.
8 atomic answers
Fleet fixture [supported] Board lists tickets across projects with status groups. Sources: looptroop:frontend.
Parent child [partial] Bead lists within a ticket; Manual QA child tickets. Sources: looptroop:frontend.
Dependencies [partial] Bead dependencies shown within a ticket only. Sources: looptroop:beads.
Attention [supported] Needs Input column and blue/red badges. Sources: looptroop:ticketflow.
History [supported] Phase navigator and logs. Sources: looptroop:frontend.
Workload [partial] Progress ring and bead counts; ETA computed from history. Sources: looptroop:frontend.
Navigation errors [supported] Errors filter and group. Sources: looptroop:frontend.
One fixed workflow with per-ticket model/effort choices and prompt edits; changes at gates only; UI exposes council and bead details live and afterwards.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
quick_task: Same full pipeline for every ticket; Manual QA optional.
growing_epic: No growth path from task to epic.
presets: Profile defaults and saved filter presets; no workflow presets.
natural_language: No natural-language workflow configuration.
in_run_changes: Edits only at gates; retry with note.
auth_friction: Requires OpenCode provider setup and gh auth.
10 atomic answers
Quick task [partial] Same full pipeline for every ticket; Manual QA optional. Sources: looptroop:ticketflow.
Growing epic [unsupported] No growth path from task to epic. Sources: looptroop:ticketflow.
Executes on the machine running the daemon (local, or a VM/dev machine the user provisions); no vendor cloud. Docker image and remote-origin settings are documented.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace), looptroop CLI (daemon control, doctor, setup, clean) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
byo_remote: README recommends a disposable VM or cloud dev machine and docs describe remote HTTPS access; no installer.
managed_cloud: No hosted offering.
hybrid: No hybrid mode.
provisioning: User provisions any VM; none by the product.
resources: Resource sizing is not specified by the project.
region: No region concept.
expiry: No product-managed expiry.
other_device_continuation: Another browser can sign in to the same daemon on the same host or a configured remote origin.
12 atomic answers
Local [supported] Daemon, OpenCode and worktrees run on the local machine. Sources: looptroop:readme.
Byo remote [partial] README recommends a disposable VM or cloud dev machine and docs describe remote HTTPS access; no installer. Sources: looptroop:readme, looptroop:config.
Self hosted [supported] The whole product is self-run; no vendor control plane. Sources: looptroop:readme.
Client disconnect [supported] Daemon and tickets continue after the browser disconnects. Sources: looptroop:cli, looptroop:ticketflow.
Other device continuation [partial] Another browser can sign in to the same daemon on the same host or a configured remote origin. Sources: looptroop:config, looptroop:cli.
One git worktree per ticket under <project>/.looptroop/worktrees; no container, VM or OS sandbox; execution runs allow-all as the user with a guard on GitHub and destructive git writes.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
container: No per-task container; Docker image only packages the daemon.
vm: No VM isolation; VM use is recommended to the user.
os_sandbox: No OS sandbox; agent runs with user privileges.
filesystem: Worktree separates files but the agent can read and write any user-accessible path.
process: Processes are not isolated.
network: No network restriction.
secrets: Daemon API and OpenCode passwords are stripped from child environments; provider and GitHub tokens remain available.
security_boundary: Docs state worktrees are workspace isolation, not host security.
12 atomic answers
Checkout [supported] Original checkout is left untouched. Sources: looptroop:readme.
Branch [supported] Ticket branch per worktree. Sources: looptroop:post.
Os sandbox [unsupported] No OS sandbox; agent runs with user privileges. Sources: looptroop:readme.
Filesystem [partial] Worktree separates files but the agent can read and write any user-accessible path. Sources: looptroop:permissions, looptroop:readme.
Process [unsupported] Processes are not isolated. Sources: looptroop:readme.
Secrets [partial] Daemon API and OpenCode passwords are stripped from child environments; provider and GitHub tokens remain available. Sources: looptroop:ops.
Code separation [supported] Planning phases are read-only against the repo; coding in worktree. Sources: looptroop:ticketflow.
Security boundary [unsupported] Docs state worktrees are workspace isolation, not host security. Sources: looptroop:readme.
Strong recovery model: serialized XState snapshots, OpenCode session ownership records, checkpoints per bead, fresh-session retries and Continue; crash recovery documented in a table.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
expiry: No sandbox expiry concept.
context_loss: Retries intentionally wipe conversation and keep a compact failure note.
duplicate_actions: Checkpoints avoid rerunning finalized beads; external side effects are not rolled back.
manual_repair: Blocked errors, git recovery receipts and conflict worktrees need manual repair at times.
12 atomic answers
Client disconnect [supported] SSE reconnect replays with Last-Event-ID; drafts saved on unload. Sources: looptroop:ticketflow.
Worker crash [supported] Interrupted beads are finalized from checkpoints, continued or reset safely. Sources: looptroop:beads.
Service restart [supported] Snapshot rehydrated at startup; corrupt snapshots go to BLOCKED_ERROR. Sources: looptroop:ticketflow.
Approved execution setup plan and runtime profile (tool cache, wrapper scripts, workspace inputs) prepare each worktree; approved Git-hook policy; no port allocation or service orchestration.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
snapshots: No environment snapshots.
services: No service orchestration documented.
ports: No port allocation for app processes.
secrets: Ignored/untracked non-reproducible files can be copied via approved workspace inputs; secrets guidance is partial.
duplicate_installs: Generated dependencies are excluded from workspace inputs and rebuilt per worktree.
collisions: No port collision handling.
remote_setup: No remote setup feature.
automatic_repair: Setup retries and Retry with note; failures reach BLOCKED_ERROR.
10 atomic answers
Dependencies [supported] Setup agent proposes and runs approved steps to provision toolchains. Sources: looptroop:beads.
Snapshots [unsupported] No environment snapshots. Sources: looptroop:srch-cloud.
Services [unsupported] No service orchestration documented. Sources: looptroop:beads.
Ports [unsupported] No port allocation for app processes. Sources: looptroop:beads.
Secrets [partial] Ignored/untracked non-reproducible files can be copied via approved workspace inputs; secrets guidance is partial. Sources: looptroop:beads.
Duplicate installs [partial] Generated dependencies are excluded from workspace inputs and rebuilt per worktree. Sources: looptroop:beads.
Collisions [unsupported] No port collision handling. Sources: looptroop:beads.
Remote setup [unsupported] No remote setup feature. Sources: looptroop:srch-cloud.
Enforced time and retry limits: AI response timeout 20 min, per-iteration 20 min, max bead retries 5, OpenCode retries, step cap, AI-question wait; no spend or token limits.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
token_limits: No token budget.
spend_limits: No spend cap.
8 atomic answers
Time limits [supported] AI response timeout, per-iteration timeout and setup timeout (1200 s defaults). Sources: looptroop:config.
Turn limits [supported] OpenCode max steps (default 0=OpenCode default) and bead retry caps. Sources: looptroop:config.
Rich logs and diagnostics (normal, AI, debug, SSE), blocked-error diagnostics, per-turn model usage, bead metrics and ETA; no cross-project statistics UI or usage export.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
usage_export: Plain-text log export; no usage CSV export.
project_stats: No project-level statistics view.
performance_method: Speed claims are not published; ADEMark lists no LoopTroop result.
responsiveness: 0.6.4 notes describe responsiveness fixes; no measurements in-repo.
13 atomic answers
Logs [supported] Per-ticket execution logs with ALL/SYS/AI/ERROR/DEBUG views and export. Sources: looptroop:api, looptroop:frontend.
Traces [supported] Session and tool traces under debug logs. Sources: looptroop:diag.
Per-turn cost_usd and token columns (input, output, reasoning, cache) are stored when OpenCode reports them and summarized per ticket and model with coverage counts; no budgets or alerts.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
child_cost: Per-model attribution within a ticket.
epic_cost: No epic entity.
harness_aggregation: Single harness aggregation across models.
billed_vs_estimated: Whether provider values are billed or estimated depends on OpenCode data and is not labelled.
subscription_allowance: No subscription allowance display.
provider_reconciliation: No reconciliation with provider invoices.
compute_cost: No compute fees.
refresh_delay: Rows are written as turns complete and broadcast over SSE.
alerts: No budget alerts.
export: Log export only.
14 atomic answers
Task cost [supported] AI details summary per ticket. Sources: looptroop:metrics.
Session cost [supported] Rows keyed by session and message. Sources: looptroop:db.
Child cost [partial] Per-model attribution within a ticket. Sources: looptroop:metrics.
Epic cost [unsupported] No epic entity. Sources: looptroop:db.
Harness aggregation [partial] Single harness aggregation across models. Sources: looptroop:metrics.
Billed vs estimated [unknown] Whether provider values are billed or estimated depends on OpenCode data and is not labelled. Sources: looptroop:metrics.
Tokens [supported] Input, output, reasoning and cache tokens. Sources: looptroop:metrics.
Subscription allowance [unsupported] No subscription allowance display. Sources: looptroop:metrics.
Provider reconciliation [unsupported] No reconciliation with provider invoices. Sources: looptroop:metrics.
Compute cost [unsupported] No compute fees. Sources: looptroop:metrics.
Refresh delay [partial] Rows are written as turns complete and broadcast over SSE. Sources: looptroop:metrics.
No compute-aware scheduling: serial beads, one execution ticket per project, bounded timeouts; no CPU/RAM limits or check reuse.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
queues: No work queue.
cpu_ram: No CPU/RAM limits.
backpressure: Rate limits on API and SSE admission.
host_awareness: No host load awareness.
check_reuse: No check reuse.
check_deduplication: No dedupe.
bounded_tests: Command timeouts and receipts bound final tests.
fairness: No fairness model.
shared_limits: Per-project lock shared by tickets.
12 atomic answers
Concurrency [supported] Execution band limit 1 per project. Sources: looptroop:ticketflow.
Single-user local tool: no presence, comments or co-prompting; others could use the same daemon token but there is no multi-user model.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
visibility: Anyone with the signed-in link or token sees the same board.
discussion: No discussion; comments are roadmap-only.
comments: No comments.
presence: No presence.
coprompting: No shared prompting.
ordering: Not applicable.
conflicts: Hash-guarded saves reject stale edits.
attribution: No user attribution.
permissions: No permissions.
11 atomic answers
Visibility [partial] Anyone with the signed-in link or token sees the same board. Sources: looptroop:cli.
Discussion [unsupported] No discussion; comments are roadmap-only. Sources: looptroop:roadmapdoc.
No organization, project or task roles, SSO or provisioning; access is a local token or cookie.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
organization_roles: No accounts, roles or SSO exist; single-user local daemon.
project_roles: No accounts, roles or SSO exist; single-user local daemon.
task_roles: No accounts, roles or SSO exist; single-user local daemon.
membership: No accounts, roles or SSO exist; single-user local daemon.
sso: No accounts, roles or SSO exist; single-user local daemon.
provisioning: No accounts, roles or SSO exist; single-user local daemon.
revocation: Sessions expire after 12 hours; token rotates per daemon start.
9 atomic answers
Organization roles [unsupported] No accounts, roles or SSO exist; single-user local daemon. Sources: looptroop:srch-team.
Project roles [unsupported] No accounts, roles or SSO exist; single-user local daemon. Sources: looptroop:srch-team.
Task roles [unsupported] No accounts, roles or SSO exist; single-user local daemon. Sources: looptroop:srch-team.
Membership [unsupported] No accounts, roles or SSO exist; single-user local daemon. Sources: looptroop:srch-team.
Sso [unsupported] No accounts, roles or SSO exist; single-user local daemon. Sources: looptroop:srch-team.
Provisioning [unsupported] No accounts, roles or SSO exist; single-user local daemon. Sources: looptroop:srch-team.
Human gates at planning artifacts, setup plan, Manual QA and PR merge; coding runs in YOLO allow-all mode with a guard on GitHub writes and destructive git.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
command: Commands are auto-allowed in execution.
tool: Tool permissions are allow-all.
network: No network approval.
secret: No secret approval.
approver_identity: Single local user; no identity recorded.
Durable per-ticket history: status history, error occurrences, edit receipts, skip trail, logs and AI details; local mutable files with no audit-integrity claim.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
discussion: No discussion.
cross_user_persistence: No users.
export: Log export.
mutability: Local files are editable by the user; no immutability.
retention: No retention policy beyond ticket deletion.
integrity: No integrity or tamper evidence model.
11 atomic answers
Actions [supported] Status history and logs. Sources: looptroop:db.
Diffs per bead and the final net diff, final test and audit summaries, candidate-file audit and PR review gate, with Manual QA checklists; no inline review comments.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
No colleague takeover model; a second person on the same host could open the ticket via the same daemon but no accounts, attribution or handoff exist.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
ongoing_takeover: No takeover feature.
same_task: Same daemon exposes same ticket.
permissions: No permissions.
runtime_context: Worktree persists on host.
ownership: No ownership transfer.
attribution: No attribution.
context_loss: Not tested.
harness_transfer: One harness; model slots fixed at Start.
12 atomic answers
Ongoing takeover [unsupported] No takeover feature. Sources: looptroop:srch-team.
Same task [partial] Same daemon exposes same ticket. Sources: looptroop:cli.
Per-ticket retained diffs, bead commits, final-test and audit reports, PR report, Manual QA evidence uploads and logs; no screenshot or recording capture by the product and no cross-task gallery.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
screenshots: Manual QA evidence uploads of images.
recordings: Manual QA accepts media evidence files.
Final test gate is hard-gated before integration; bead completion requires declared test, lint and typecheck results; Git-hook policy can require validation commands; independent council review and coverage loops; no CI gate or spec-conformance across tasks.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
lint: Declared lint outcomes and optional hook validation.
types: Declared typecheck outcomes.
ci: No CI integration or wait.
independent_review: Council coverage verification and a candidate-file audit; no independent code reviewer.
spec_checks: Coverage checks between PRD and beads.
requirement_conflicts: Within-ticket only.
11 atomic answers
Tests [supported] Final test plan runs and gates integration; bead tests via BEAD_STATUS. Sources: looptroop:post, looptroop:beads.
Independent review [partial] Council coverage verification and a candidate-file audit; no independent code reviewer. Sources: looptroop:llm, looptroop:post.
Spec checks [partial] Coverage checks between PRD and beads. Sources: looptroop:ticketflow.
Integrated git workflow: per-ticket worktree and branch, per-bead commits, squash to one candidate, force-with-lease push and draft PR via `gh`, merge verification and conflict recovery receipts.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
comments: No PR comment handling.
ci_response: No CI response.
rebase: Early-merge recovery re-applies deltas on verified base.
User merges via the Merge button (GitHub merge commit) or on GitHub; app verifies head and landed commit; no release or deploy authority.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
release: No release feature.
deploy: No deploy feature.
authority_enforcement: Uses the user gh credentials; no separate authority model.
Ticket links to candidate SHA, PR URL and merge report with remote landed-commit proof; no builds, deployments or automated rollback.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
build_links: No builds.
deployment_links: No deployments.
rollback: No rollback.
revert: No revert action.
9 atomic answers
Commit links [supported] Candidate and bead commits. Sources: looptroop:post.
Pr links [supported] PR URL and number. Sources: looptroop:post.
Per-ticket review of a squashed PR with audits; no batch review queue across tasks, dependency-aware merge queue or risk filters.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
review_queue: Needs Input column and PR review status.
grouping: No grouping.
risk_size_filters: Ticket size buckets exist; no risk filters.
request_changes: Manual QA failure loop.
dependency_queues: No dependency-aware queue across tickets.
MIT for the single application repository (daemon, CLI and UI); docs live in a separate public website repo; no hosted component or paid extension.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Self-hosting is the product: local daemon with SQLite and files in the project and a config directory; no vendor account or service. Artifacts are YAML/JSONL; no export/import tooling.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
vendor_dependencies: Needs GitHub, `gh` and OpenCode providers; no LoopTroop vendor service (update check only).
export: Plain-text log export and readable .ticket YAML/JSONL; backup is copying directories.
import: No import.
history_portability: Files and DB are copyable; version 0.6.4 refuses older databases.
All state stays local; prompts and code go to the model providers configured in OpenCode; only outbound LoopTroop call found is the GitHub release check; no telemetry code found.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Free and open source; no plans or prices. Users pay model providers and their own machines.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
currency: No price is published.
interval: No billing.
seats: No seats.
seat_minimum: None.
workspace_fees: None.
included_usage: None; usage is billed by providers.
credits: No vendor credits.
trial: No trial needed.
overages: None.
taxes: Not applicable.
12 atomic answers
Currency [unknown] No price is published. Sources: looptroop:website.
No scenario pricing: the software is free; costs are provider tokens and the host machine, which the project says can be high for councils and retry loops.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
one_person: Software cost zero; model cost depends on provider and is high-volume per README.
two_people: Same; no collaboration features.
five_people: Same.
task_assumptions: No standard scenario is published.
concurrency: Not specified.
tokens: No average token figures are published.
compute_hours: Not specified.
variable_cost: Tokens and machine time.
included_cost: Nothing included.
11 atomic answers
One person [partial] Software cost zero; model cost depends on provider and is high-volume per README. Sources: looptroop:readme.
Two people [partial] Same; no collaboration features. Sources: looptroop:readme.
Generally available as early alpha; 27 releases since April 2026, latest 2026-11-06; GitHub issues and advisories for support; best-effort responses; latest-only security fixes.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
release_stage: Early alpha per README; releases are public.
support: GitHub issues and private security advisories.
response_commitments: Security policy says best effort, no contractual timelines.
Free forever as open source (MIT); no card, caps or trial; real work needs a configured OpenCode provider whose models cost money unless free-tier models are used.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
trial: No trial needed.
credits_trial: No credits.
paid_only: Not paid-only.
early_access: Public.
free_scenarios: Single-user local, BYO subscriptions or free models, no collaboration, mobile or remote features.
card_required: No payment method.
usage_caps: No product caps; provider rate limits apply.
paid_evaluation: Evaluation can use free OpenRouter models; stronger models need paid access.
Not applicable beyond free software: no pricing page; costs are third-party model and machine charges.
Scope
Surfaces: Local browser UI served by the LoopTroop daemon (Kanban board and ticket workspace) · Version: v0.6.4 (commit cf617d0, released 2026-11-06); docs at LoopTroop-Website commit 3368cb4 · Plan: Open source (MIT), free; self-run on user machine · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
generally-available
Delivery
native
Checked
2026-11-17 (5 days old)
Agent cross-check
2026-11-17
units: No pricing units.
included_limits: None.
overages: None.
compute_markup: No compute resale.
model_markup: No model resale.
minimum_commitment: None.
cancellation: No subscription.
sales_only_unknowns: No sales-only tier.
transparency_classification: The catalog classification needs a pricing page; none exists for this free tool.
Daemon binds 127.0.0.1:3000, keeps running after the shell closes, `open` signs the browser in with a single-use code; 12-hour sessions; bearer token for automation.
curl/PowerShell, npm, Homebrew, Scoop, Chocolatey, WinGet, bun, pnpm, Yarn, Docker and standalone executables for macOS arm64, Linux x64/arm64 and Windows x64.
Kind
documented
Locator
Channels table; Platforms with no executable; Running in a container
Ticket state machine from DRAFT to COMPLETED with approval gates, a Kanban board, one ticket per project in the execution band, BLOCKED_ERROR recovery and restart rehydration.
Kind
documented
Locator
At A Glance; Execution Band; Kanban Board Locations; User Actions; Safe Resume
Needs OpenCode with a configured provider and a local git repo with a GitHub origin; models and subscriptions come from OpenCode; free OpenRouter, NVIDIA NIM and OpenCode Go options.
Kind
documented
Locator
Prerequisites; Setting Up Your AI Council; Free Model Options; Attaching Your First Project
LoopTroop supervises or adopts an OpenCode server, uses its providers, MCP servers and skills, owns sessions, guards direct GitHub and destructive git writes and reconnects after restarts.
Kind
documented
Locator
Core Modules; Session Ownership; Tool Policy Layer; Reconnect Behavior; Questions And Human Input
Local GUI orchestrator: council planning, beads, Ralph-style retries, OpenCode execution in git worktrees, human approval gates and a PR handoff; MIT; early alpha; recommends running in a disposable VM because OpenCode runs in YOLO mode.
Kind
documented
Locator
Quick start; How it works; Worktree isolation; Run it in a VM; Project status
Final test gate, optional Manual QA, squash into one candidate commit, draft PR via GitHub, human Merge or Finish Without Merge, cleanup that preserves evidence.
Local REST API under /api with cookie or bearer authentication, ticket lifecycle actions, artifacts, logs export, prompt editing and SSE streams; same API as the UI.
Beads carry dependencies.blocked_by, priority, target files and tests; scheduler runs only pending beads whose blockers are done; retries use fresh sessions with a context-wipe note.
Records cost_usd, input/output/reasoning/cache tokens and duration per completed assistant message and summarises them per ticket with coverage counts.
Only outbound product call found: GET to api.github.com releases/latest every 15 minutes at most to show update steps; nothing is installed or sent about the user.