51 of 51 criteria have a sourced finding; 48 were also cross-checked by a second agent. Every comparison that includes stagewise 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.
Electron desktop app for macOS (arm64, x64), Windows x64 and Linux x64 (deb/rpm); no web, mobile or editor-extension client. A headless CLI exists only as a private MVP.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent), stagewise-cli headless MVP host (private workspace package, not distributed) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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
No browser, mobile or editor-extension client.
Headless CLI is a private MVP.
Minimum OS versions unstated.
9 atomic answers
Desktop os [supported] Release workflow builds macOS arm64/x64, Linux x64 and Windows x64 installers (DMG/ZIP, deb/rpm, Squirrel); minimum OS versions are not stated in the source. Sources: stagewise:release-wf, stagewise:forge.
Browser [unsupported] The product is an Electron app; apps/website is a marketing site and no web client for agent work exists in the repository. Sources: stagewise:agents-md, stagewise:release-wf.
Mobile browser [unsupported] No web client exists, so no phone-browser access. Sources: stagewise:agents-md.
Editor extensions [unsupported] Bounded search for vscode extension, vsix and cursor extension terms outside the website finds none; the app is the editor. Sources: stagewise:agents-md.
Built-in agent plus external Claude Code, Codex and OpenCode sessions through ACP, and a long list of model providers including custom endpoints.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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
custom_adapters: partial
mixed_harness_operation: partial
5 atomic answers
Harnesses [supported] Built-in stage agent and ACP adapters for claude-code, codex and opencode. Sources: stagewise:acp, stagewise:registry.
Model providers [supported] Anthropic, OpenAI, Google, Moonshot, Alibaba, DeepSeek, Z.ai, MiniMax, Xiaomi, Mistral, xAI, Azure, Bedrock, Vertex, OpenRouter, Ollama and custom compatible endpoints are registered. Sources: stagewise:registry.
Custom adapters [partial] Custom Anthropic/OpenAI/Google-compatible model endpoints are supported, but the ACP harness registry is a fixed typed set with no user-defined adapter. Sources: stagewise:registry, stagewise:acp.
Session integration [supported] The built-in agent runs in-process; external agents are controlled over ACP sessions with resume support. Sources: stagewise:acp-runtime, stagewise:agent-manager.
Mixed harness operation [partial] Chats can use different harnesses (built-in and Claude Code/Codex/OpenCode) in one workspace, but each chat uses one harness; no cross-harness team. Sources: stagewise:acp, stagewise:agent-manager.
Subscriptions through Claude Code login, ChatGPT login via Codex, and imported coding plans; own API keys; local models; no vendor account needed, while stagewise Cloud Inference credits are being sunset.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Api accounts [supported] Own API keys for the registered provider types and custom endpoints. Sources: stagewise:registry, stagewise:readme.
Vendor credits [partial] stagewise account plans include Cloud Inference credits, but the account service is announced to sunset after November 2026. Sources: stagewise:pricing, stagewise:readme.
Account switching [supported] Several provider instances (including multiple of one type) can be configured and models chosen per chat. Sources: stagewise:provider-model.
Vendor account requirement [supported] No stagewise account is needed since 1.31.0 removed account-first onboarding; BYO keys or subscriptions work. Sources: stagewise:changelog, stagewise:readme.
Provider restrictions [unknown] Provider terms for third-party subscription reuse were not assessed in first-party stagewise documents.
Download the app, complete onboarding (providers, skills, theme), mount a folder and chat; git is optional, and there is no managed-cloud or VPS path.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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
No managed cloud or BYO VPS setup.
Prerequisite guidance is mostly in-app errors.
7 atomic answers
First task setup [supported] Onboarding walks through provider configuration, skills and theme; a task starts by mounting a workspace and chatting. Sources: stagewise:onboarding, stagewise:readme.
No git [supported] Workspaces mount as folders; git actions report not-git-repo rather than blocking use. Sources: stagewise:git-types, stagewise:mount.
Local git without remote [supported] Worktree and branch actions are local git operations; remote fetch is optional. Sources: stagewise:git.
Hosted git [supported] A clone of a hosted repository works as a normal local repository; there is no hosted-git integration beyond optional GitHub token plugin. Sources: stagewise:git, stagewise:credentials.
Prerequisite guidance [partial] External agents error with install/sign-in instructions; the README gives a short getting-started pointer to the download page. Sources: stagewise:claude-code, stagewise:readme.
Managed cloud setup [unsupported] No managed agent execution; Cloud Inference only routes model calls. Sources: stagewise:klex-news, stagewise:site.
Byo vps onboarding [unsupported] Agents run locally; the vendor post says remote execution is not offered. Sources: stagewise:klex-news.
Api [unsupported] Bounded search for webhook, REST/server entry points and published API docs finds none beyond Electron IPC. Sources: stagewise:agents-md.
Sdk [unsupported] No SDK for controlling stagewise is published; the karton package is a state-transport library. Sources: stagewise:agents-md.
Mcp server [unsupported] The only MCP server is an internal stdio helper given to external ACP agents for user questions; it exposes no create/steer/review tools. Sources: stagewise:mcp.
Webhooks [unsupported] No webhook code found in apps/browser or packages. Sources: stagewise:agents-md.
Operations [partial] CLI can start a prompt and print the result; stop, inspect and review are not scriptable. Sources: stagewise:cli-readme.
Authentication [unsupported] No product API exists to authenticate; the CLI reads a provider API key from the environment. Sources: stagewise:cli-readme.
Stability [unsupported] The only programmatic entry is a private MVP with no stability promise. Sources: stagewise:cli-pkg.
No user-configurable MCP client; extensions come from skills and bundled credential-backed plugins, and instructions import from AGENTS.md and several skill folders.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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: unsupported
plugins: partial
custom_tools: unsupported
permission_scope: partial
7 atomic answers
Mcp client [unsupported] ACP sessions receive only the internal stagewise MCP server and no user MCP configuration surface was found in the pinned release. Sources: stagewise:acp-runtime, stagewise:mcp.
Plugins [partial] Bundled plugins (Figma, GitHub, Vercel, Supabase, PostHog, Remotion and others) are skill directories with metadata; no third-party install flow found. Sources: stagewise:bundled, stagewise:plugins.
Custom tools [unsupported] Tools are built into the toolbox; no user tool registration found. Sources: stagewise:bundled.
Instructions import [supported] Reads workspace AGENTS.md and skills from .stagewise/.agents/.claude/.codex skill folders. Sources: stagewise:skills, stagewise:agentsmd.
Authentication [supported] A credential registry stores personal access tokens for plugin services. Sources: stagewise:credentials.
Permission scope [partial] Each plugin lists required credentials; token scope is the user-created token scope. Sources: stagewise:plugins, stagewise:credentials.
Integration owner [supported] Plugin integrations are product-managed; Claude Code/Codex keep their own harness tooling. Sources: stagewise:bundled, stagewise:claude-code.
Electron with React 19 for the desktop client; no mobile client.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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: unsupported
4 atomic answers
Desktop implementation [supported] Electron with React 19 and Tailwind; Electron Forge builds installers. Sources: stagewise:agents-md, stagewise:forge.
Mobile implementation [unsupported] No mobile client exists in the repository. Sources: stagewise:agents-md.
Hybrids [supported] Single Electron web-technology hybrid with a Node backend and embedded Chromium pages. Sources: stagewise:agents-md.
Prompt intake [supported] Prompt intake through the chat input with attachments. Sources: stagewise:notify.
External import [partial] No issue-import feature, but the bundled GitHub plugin skill lets the agent read and search GitHub issues with a PAT; no Linear or Jira integration exists in the app source (only an icon component and Klex Bots marketing). Sources: stagewise:github-plugin, stagewise:klex-news.
Chat git triggers [unsupported] No webhook, Slack or git-event trigger code found. Sources: stagewise:agents-md.
Authoritative tracker [unsupported] No tracker relationship; the chat is the unit of work. Sources: stagewise:history-ui.
Stable ids [partial] Chats have persisted agent IDs but no human-facing issue identifiers. Sources: stagewise:persistence.
A bundled plan skill writes markdown plans shown in a plan viewer; there is no acceptance-criteria model or revision history beyond chat and diff history.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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
acceptance_criteria: unsupported
editors: partial
revisions: partial
implementation_links: unsupported
7 atomic answers
Plans [supported] Plans are markdown documents parsed into sections; bundled plan skill creates them. Sources: stagewise:plans, stagewise:bundled.
Acceptance criteria [unsupported] No acceptance-criteria structure found in plan parsing. Sources: stagewise:plans.
Editors [partial] Plans are files the user can edit in the mounted workspace; the plan viewer is read oriented. Sources: stagewise:plans.
Revisions [partial] Edits are tracked by diff history; plans have no dedicated revision list. Sources: stagewise:diff-history.
Implementation links [unsupported] No link between plan items and commits or tasks found. Sources: stagewise:plans.
Amendment [supported] Messages can be queued and sent while an agent runs, and plan files can be changed by the agent. Sources: stagewise:notify.
No living product specification feature; skills and AGENTS.md are static instruction files.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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
9 atomic answers
Persistent requirements [unsupported] Bounded search of agent-core and apps/browser for spec/requirement stores finds none; only static AGENTS.md and skills are read. Sources: stagewise:agentsmd, stagewise:skills.
Human edits [unsupported] Bounded search of agent-core and apps/browser for spec/requirement stores finds none; only static AGENTS.md and skills are read. Sources: stagewise:agentsmd, stagewise:skills.
Agent edits [unsupported] Bounded search of agent-core and apps/browser for spec/requirement stores finds none; only static AGENTS.md and skills are read. Sources: stagewise:agentsmd, stagewise:skills.
Revisions [unsupported] Bounded search of agent-core and apps/browser for spec/requirement stores finds none; only static AGENTS.md and skills are read. Sources: stagewise:agentsmd, stagewise:skills.
Approvals [unsupported] Bounded search of agent-core and apps/browser for spec/requirement stores finds none; only static AGENTS.md and skills are read. Sources: stagewise:agentsmd, stagewise:skills.
Implementation links [unsupported] Bounded search of agent-core and apps/browser for spec/requirement stores finds none; only static AGENTS.md and skills are read. Sources: stagewise:agentsmd, stagewise:skills.
Later retrieval [unsupported] Bounded search of agent-core and apps/browser for spec/requirement stores finds none; only static AGENTS.md and skills are read. Sources: stagewise:agentsmd, stagewise:skills.
Drift detection [unsupported] Bounded search of agent-core and apps/browser for spec/requirement stores finds none; only static AGENTS.md and skills are read. Sources: stagewise:agentsmd, stagewise:skills.
Review integration [unsupported] Bounded search of agent-core and apps/browser for spec/requirement stores finds none; only static AGENTS.md and skills are read. Sources: stagewise:agentsmd, stagewise:skills.
Quotas [unsupported] No concurrency limit or quota found in agent-manager. Sources: stagewise:agent-manager.
Observed concurrency [partial] Code supports concurrency; the catalog does not run the product. Sources: stagewise:agent-manager.
Status visibility [partial] Sidebar and notifications show running, question and error states; no board. Sources: stagewise:history-ui, stagewise:notify.
No schedules or event triggers; work starts only when a user sends a message.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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
8 atomic answers
Schedules [unsupported] Bounded search for cron, scheduled task and webhook code in apps/browser and packages finds none; the vendor says local agents cannot work 24/7. Sources: stagewise:klex-news, stagewise:agent-manager.
Events [unsupported] Bounded search for cron, scheduled task and webhook code in apps/browser and packages finds none; the vendor says local agents cannot work 24/7. Sources: stagewise:klex-news, stagewise:agent-manager.
Recurring tasks [unsupported] Bounded search for cron, scheduled task and webhook code in apps/browser and packages finds none; the vendor says local agents cannot work 24/7. Sources: stagewise:klex-news, stagewise:agent-manager.
Completion conditions [unsupported] Bounded search for cron, scheduled task and webhook code in apps/browser and packages finds none; the vendor says local agents cannot work 24/7. Sources: stagewise:klex-news, stagewise:agent-manager.
Scheduler location [unsupported] Bounded search for cron, scheduled task and webhook code in apps/browser and packages finds none; the vendor says local agents cannot work 24/7. Sources: stagewise:klex-news, stagewise:agent-manager.
Overlaps [unsupported] Bounded search for cron, scheduled task and webhook code in apps/browser and packages finds none; the vendor says local agents cannot work 24/7. Sources: stagewise:klex-news, stagewise:agent-manager.
Retries [unsupported] Bounded search for cron, scheduled task and webhook code in apps/browser and packages finds none; the vendor says local agents cannot work 24/7. Sources: stagewise:klex-news, stagewise:agent-manager.
Client independence [unsupported] Bounded search for cron, scheduled task and webhook code in apps/browser and packages finds none; the vendor says local agents cannot work 24/7. Sources: stagewise:klex-news, stagewise:agent-manager.
Escalation [partial] Question and error events notify the user; no escalation chain. Sources: stagewise:notify.
Visible errors [supported] Error events surface in chat and notifications. Sources: stagewise:notify.
Chat terminal parity [partial] External CLI agents run through ACP rather than their own terminal UI, so CLI-only questions use the stagewise form bridge. Sources: stagewise:mcp, stagewise:acp-runtime.
Retained history [supported] History persists locally. Sources: stagewise:persistence.
Git worktrees and branches created from the chat, a JS sandbox utility process and permissioned mounts; shell commands run as the user with approvals; no container or VM.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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
container: unsupported
vm: unsupported
os_sandbox: unsupported
filesystem: partial
12 atomic answers
Checkout [supported] Agents can work in the existing checkout. Sources: stagewise:mount.
Branch [supported] Branch creation and switching. Sources: stagewise:git.
Os sandbox [unsupported] Shell runs as the user; the JS sandbox is a utility process, not an OS sandbox. Sources: stagewise:sandbox, stagewise:shell-tool.
Filesystem [partial] Mount permission flags scope tool file access; shell can reach beyond them with approval. Sources: stagewise:mount, stagewise:smart-approval.
Process [partial] Agent shell commands are ordinary user processes. Sources: stagewise:shell-tool.
maxSteps and maxRetries agent config, model fallback on overload and tool-approval gating; no time, token or spend caps.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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: unsupported
turn_limits: unsupported
token_limits: unsupported
spend_limits: unsupported
8 atomic answers
Time limits [unsupported] No wall-clock limits found. Sources: stagewise:notify.
Turn limits [unsupported] BaseAgent has an optional maxSteps config, but a search of apps/browser/src/backend and the chat agent finds nothing that sets it. Sources: stagewise:notify.
Token limits [unsupported] No token caps found. Sources: stagewise:notify.
Spend limits [unsupported] No spend limits found. Sources: stagewise:usage.
Timeouts [partial] The shell service and schemas carry command timeouts (agent-shell); provider-call timeouts use SDK defaults. Sources: stagewise:shell-tool.
No compute-aware scheduling; concurrency is bounded only by the machine, with bounded-concurrency helpers for model calls.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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: unsupported
queues: unsupported
cpu_ram: unsupported
backpressure: unsupported
12 atomic answers
Concurrency [unsupported] No scheduler, resource limits or check deduplication found in agent-manager or apps/browser. Sources: stagewise:agent-manager, stagewise:shell-tool.
Backpressure [unsupported] No scheduler, resource limits or check deduplication found in agent-manager or apps/browser. Sources: stagewise:agent-manager, stagewise:shell-tool.
Host awareness [unsupported] No scheduler, resource limits or check deduplication found in agent-manager or apps/browser. Sources: stagewise:agent-manager, stagewise:shell-tool.
Check reuse [unsupported] No scheduler, resource limits or check deduplication found in agent-manager or apps/browser. Sources: stagewise:agent-manager, stagewise:shell-tool.
Check deduplication [unsupported] No scheduler, resource limits or check deduplication found in agent-manager or apps/browser. Sources: stagewise:agent-manager, stagewise:shell-tool.
Bounded tests [unsupported] No scheduler, resource limits or check deduplication found in agent-manager or apps/browser. Sources: stagewise:agent-manager, stagewise:shell-tool.
Shared limits [unsupported] No scheduler, resource limits or check deduplication found in agent-manager or apps/browser. Sources: stagewise:agent-manager, stagewise:shell-tool.
Single-user local app: no shared sessions, comments, presence or co-prompting.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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
11 atomic answers
Visibility [unsupported] Bounded search for collaborator, multiplayer and presence code finds only unrelated matches; chats are stored locally. Sources: stagewise:persistence.
Discussion [unsupported] Bounded search for collaborator, multiplayer and presence code finds only unrelated matches; chats are stored locally. Sources: stagewise:persistence.
Comments [unsupported] Bounded search for collaborator, multiplayer and presence code finds only unrelated matches; chats are stored locally. Sources: stagewise:persistence.
Presence [unsupported] Bounded search for collaborator, multiplayer and presence code finds only unrelated matches; chats are stored locally. Sources: stagewise:persistence.
Coprompting [unsupported] Bounded search for collaborator, multiplayer and presence code finds only unrelated matches; chats are stored locally. Sources: stagewise:persistence.
Ordering [unsupported] Bounded search for collaborator, multiplayer and presence code finds only unrelated matches; chats are stored locally. Sources: stagewise:persistence.
Conflicts [unsupported] Bounded search for collaborator, multiplayer and presence code finds only unrelated matches; chats are stored locally. Sources: stagewise:persistence.
Attribution [unsupported] Bounded search for collaborator, multiplayer and presence code finds only unrelated matches; chats are stored locally. Sources: stagewise:persistence.
Permissions [unsupported] Bounded search for collaborator, multiplayer and presence code finds only unrelated matches; chats are stored locally. Sources: stagewise:persistence.
Plan restrictions [unsupported] Bounded search for collaborator, multiplayer and presence code finds only unrelated matches; chats are stored locally. Sources: stagewise:persistence.
Owner online [unsupported] Bounded search for collaborator, multiplayer and presence code finds only unrelated matches; chats are stored locally. Sources: stagewise:persistence.
No roles in the open-source app; the Enterprise account plan advertises SSO (OIDC, SAML) and SCIM on the pricing page, which the code does not implement.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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: unsupported
project_roles: unsupported
task_roles: unsupported
membership: unknown
9 atomic answers
Organization roles [unsupported] No role code in apps/browser or packages. Sources: stagewise:persistence.
Tool approvals gate shell and sandbox commands in always-ask (default), smart (LLM classifier) or allow-all modes; no network, secret or delivery approvals.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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
tool: partial
network: partial
secret: unsupported
delivery: unsupported
14 atomic answers
Command [supported] Shell commands need approval by default. Sources: stagewise:approval.
Branches and worktrees are product-managed in the UI; commits and rebase run through ordinary shell git, and the bundled GitHub plugin skill lets the agent create, review, comment on and merge pull requests through the GitHub REST API with a user PAT (no dedicated PR UI).
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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
Not separately cross-checked
commits: unsupported
pull_requests: unsupported
comments: unsupported
ci_response: unsupported
9 atomic answers
Branches [supported] Create and switch branches. Sources: stagewise:git.
Commits [unsupported] No product commit action; no commit tool found. Sources: stagewise:git.
Pull requests [partial] No native PR button or view, but the bundled GitHub plugin skill documents creating, listing, reviewing and merging pull requests through the REST API using a stored PAT. Sources: stagewise:github-plugin, stagewise:credentials, stagewise:bundled.
Comments [partial] The bundled GitHub plugin skill documents commenting on and reviewing pull requests via the REST API; there is no native comment UI. Sources: stagewise:github-plugin.
Ci response [unsupported] None. Sources: stagewise:git.
Rebase [unsupported] No product rebase. Sources: stagewise:git.
Merge [partial] No product merge control or gate; the bundled GitHub plugin skill documents merging pull requests through the REST API when the agent uses it. The app also detects merged worktree targets. Sources: stagewise:github-plugin, stagewise:git.
AGPL-3.0 for the repository (stagewise GmbH), with contact offered for uses outside AGPL; hosted Cloud Inference is closed.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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: partial
hosted_extensions: unknown
hosted_parity: unknown
6 atomic answers
Client license [supported] GNU AGPL v3 license file; GitHub reports AGPL-3.0; the file states version 3 without later-version grant. Sources: stagewise:license, stagewise:repo.
Server license [partial] No separate server in the repository; hosted stagewise Cloud is closed and sunsetting. Sources: stagewise:readme.
Hosted extensions [unknown] Hosted Cloud Inference and enterprise features are closed.
Source visibility [supported] Public monorepo. Sources: stagewise:repo.
The whole app runs locally with BYO keys; stagewise accounts are optional and being shut down; chat history is local libsql with no documented export.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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
export: unsupported
import: unsupported
formats: partial
cancellation: unknown
8 atomic answers
Self hosting [supported] The desktop app runs on user machines; Enterprise lists self-hosting the stagewise Cloud for inference. Sources: stagewise:readme, stagewise:pricing.
Vendor dependencies [supported] No required vendor service when using BYO keys. Sources: stagewise:changelog.
Code goes to the chosen model provider; stagewise account routing uses Vercel AI Gateway without zero-retention guarantee; PostHog telemetry has off, anonymous (default) and full levels.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · 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: partial
logs: partial
deletion: partial
training: partial
15 atomic answers
Code data [supported] Customer content processed ephemerally on account routing; with BYO keys requests go to the chosen provider. Sources: stagewise:privacy, stagewise:registry.
Credentials [partial] Provider keys are stored encrypted in preferences. Sources: stagewise:provider-model.
Logs [partial] Local logs; telemetry events go to PostHog. Sources: stagewise:telemetry.
Artifacts [supported] Multimodal assets stored up to 72 hours on account routing. Sources: stagewise:privacy.
Retention [supported] Policy states 72 hours for multimodal assets and no permanent codebase storage. Sources: stagewise:privacy.
Deletion [partial] GDPR erasure requests are described. Sources: stagewise:privacy.
Training [partial] Policy states no zero-data-retention guarantee from AI providers and reserves the right to use collected data to improve the service; it makes no explicit training statement. Sources: stagewise:privacy.
Application is free; the stagewise account listing shows Hobby free, Individual $20/month and Enterprise custom, but account service sunsets after November 2026; model costs are separate.
Scope
Surfaces: stagewise account and Cloud Inference (hosted, being sunset) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: stagewise account (Hobby, Individual, Enterprise) · 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
seats: partial
seat_minimum: unknown
workspace_fees: unsupported
included_usage: partial
12 atomic answers
Currency [supported] Individual priced in USD. Sources: stagewise:pricing.
Interval [supported] Individual is monthly; no annual option listed. Sources: stagewise:pricing.
Seats [partial] Individual is per account; Enterprise custom. Sources: stagewise:pricing.
Seat minimum [unknown] Not stated.
Workspace fees [unsupported] No workspace fee on the pricing page. Sources: stagewise:pricing.
Included usage [partial] Hobby includes limited access to three models; amounts unstated. Sources: stagewise:pricing.
Credits [partial] Individual includes extra credits to extend usage; amounts unstated. Sources: stagewise:pricing.
Trial [unknown] No trial described.
Overages [partial] Credits can be purchased; rates unstated. Sources: stagewise:pricing.
GA desktop releases with frequent minor releases through 1.31.1 (2026-11-04); vendor says development is mostly halted and only fixes and small improvements continue; support through GitHub issues and Discord.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Open-source app is free to run with BYO keys or a free Hobby account (service being sunset); no card needed to use the app.
Scope
Surfaces: stagewise desktop app (Electron agentic IDE with built-in browser and agent), stagewise account and Cloud Inference (hosted, being sunset) · Version: stagewise@1.31.1 (commit 17a0123a0e896885465df36169457199d5225eb5, 2026-11-04); source pass 2026-11-16 · Plan: Not specified · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Early access [unsupported] No gated early access. Sources: stagewise:changelog.
Free scenarios [partial] BYO subscriptions, local models and review work free; no collaboration, remote compute, mobile or issue management exist. Sources: stagewise:readme.
Card required [unknown] Not stated.
Usage caps [partial] Hobby has limited model access; BYO is described as unlimited. Sources: stagewise:pricing.
Paid evaluation [supported] No payment needed to evaluate with BYO keys. Sources: stagewise:readme.
Monorepo layout: apps/browser (Electron main product), apps/stagewise-cli (headless host for extracted agent packages), apps/deprecated-cli (legacy v0.12), website, update-server, agent-core and agent-shell packages. Product is an Electron app.
Describes an open source agentic IDE with built-in coding agent, browser, BYO keys, listed subscription/provider imports, an account-service sunset at end of November 2026, and AGPLv3 licensing with contact for other uses.
Kind
source-inspected
Locator
sections About, Use your coding subscription, stagewise Account (sunset), License
Hobby free with limited access to three models; Individual $20/month (Pro and Ultra grouped) with all models and extra credits; Enterprise custom with SSO (OIDC, SAML), SCIM and optional self-hosted stagewise Cloud. No annual option or credit amounts stated.
Launches the user-installed claude executable through the bundled Claude agent ACP adapter; errors if claude is not installed or signed in; ask permissions map from the approval mode.
Reads the Codex CLI auth.json ChatGPT access token so the built-in stagewise agent can use a ChatGPT subscription via the Codex responses endpoint; asks the user to run codex login first.
Users hold a list of provider instances (several of the same type possible) and a privacy.telemetryLevel preference of off, anonymous or full defaulting to anonymous.
Dated 2026-10-28. Says local agents cannot work 24/7, the team mostly halted work on stagewise, stagewise IDE accounts are being sunset (BYO keys needed after November 2026) and bug fixes continue for the foreseeable future.
Kind
documented
Locator
sections Why we are building Klex Bots; What changes for stagewise
Positions stagewise as an agentic IDE and agent orchestrator running locally; stagewise Account gives preconfigured Cloud Inference; BYOK and custom endpoints supported; open source AGPLv3; no download links or docs on the page.
The only MCP surface is a private stdio server handed to external ACP agents so they can show a native question form; no user MCP server configuration or exported control tools.
Skills are mounted from ~/.stagewise/skills, ~/.agents/skills and optionally ~/.codex/skills and ~/.claude/skills; workspace skills come from .stagewise/skills and .agents/skills.
Plugins are directories with metadata.json (name, description, required credential types) plus skills; credentials come from the product credential registry.
The product-bundled GitHub plugin skill tells the agent it can create, read, update and comment on issues, and create, list, review and merge pull requests through the GitHub REST API with a stored personal access token fetched inside the JS sandbox.
Kind
source-inspected
Locator
Capabilities list (issues, pull requests, actions, releases) and Making Requests section
In smart mode an LLM classifier decides whether a shell command needs approval, flagging out-of-workspace writes, global installs, outbound data, destructive git and privilege escalation.
Effective 2026-03-09. Customer content processed ephemerally; multimodal assets stored up to 72 hours in AWS S3; requests routed through Vercel AI Gateway to named inference providers without a zero-retention guarantee; PostHog EU analytics is opt-in.
Kind
source-inspected
Locator
sections Customer Content, AI Model Interaction, Usage and Analytics, sub-processors, Data Retention