generated: '2026-09-01' method: probed source: https://api.stellary.co/mcp (live probe) + https://stellary.co/docs/mcp/ + https://registry.modelcontextprotocol.io/v0/servers?search=stellary name: Stellary Project Management MCP Server registry_id: io.github.Anymfah/stellary-project-management registry_status: active registry_url: https://registry.modelcontextprotocol.io/v0.1/servers/io.github.Anymfah%2Fstellary-project-management/versions/latest repository: https://github.com/Anymfah/stellary-mcp documentation: https://stellary.co/docs/mcp/ version: 0.13.0 version_note: >- server.json and every plugin manifest in the public discovery repo declare 0.13.0 (repo release tag 0.12.0, published 2026-08-30); the official MCP registry record still shows 0.12.0 as latest, published 2026-07-27. Both numbers read 2026-09-01. deployment: mode: remote endpoint: https://api.stellary.co/mcp auth: oauth verified: probed note: >- Remote only, and the provider says so explicitly: SKILL.md states "Not supported: local npx/stdio servers or Dockerized copies of the backend", and the README says the hosted application and server implementation are not distributed from the discovery repository. Every client manifest in that repo (.mcp.json, .cursor-plugin, .claude-plugin, .grok-plugin, gemini-extension.json, plugins/stellary/.mcp.json) points at the same hosted URL with no install command. There is no npm/PyPI package -- the only package.json in the repo is `stellary-mcp-registry-metadata`, marked private, and no `stellary*` package exists on npm or PyPI. transport: type: streamable-http methods: [GET, POST] session: stateless session_note: A fresh transport is created per request; no Mcp-Session-Id is required. authentication: primary: OAuth 2.1 (PKCE S256, dynamic client registration, rotating refresh tokens) discovery: authorization_server_metadata: https://api.stellary.co/.well-known/oauth-authorization-server protected_resource_metadata: https://api.stellary.co/.well-known/oauth-protected-resource/mcp scopes: [projects:read, projects:write, pilotage:read, pilotage:write, offline_access] access_token_lifetime: 1 hour alternatives: - Personal access token (Bearer, prefix sl_) for clients that cannot complete remote OAuth - User JWT from the /auth flows - Agent token for workspace agents running queued missions - MCP_TOKEN, a transport gate only -- not sufficient for tools that need a user or agent identity x-evidence: fetched: '2026-09-01' probe: POST https://api.stellary.co/mcp {"jsonrpc":"2.0","id":1,"method":"tools/list"} http_status: 401 body: '{"error":"Bearer token required"}' www_authenticate: >- Bearer resource_metadata="https://api.stellary.co/.well-known/oauth-protected-resource/mcp", scope="projects:read projects:write pilotage:read pilotage:write" conclusion: >- The server is live and correctly implements the MCP OAuth challenge, so tools/list cannot be introspected anonymously. The tool list below is the provider-published one from stellary.co/docs/mcp/ and the provider-authored SKILL.md -- names and descriptions only. Input schemas require an authenticated tools/list. tools_source: published (docs + provider SKILL.md); input schemas NOT available anonymously tool_groups: - name: Board read identity: core (human PAT or OAuth) tools: - name: list_projects description: List projects visible to the caller. The recommended first call after connecting. - name: get_project_details description: Read one project, including columns and delivery context. - name: list_cards description: List cards, typically filtered by column. - name: get_card_details description: Read a single card. - name: get_card_comments description: Read the comment thread on a card. - name: Board write and collaboration identity: core (write scopes) tools: - name: create_card description: Create one card. - name: create_cards_bulk description: Create several cards in one call. - name: move_card description: Move a card between columns. - name: update_card description: Update card fields. - name: assign_card description: Assign an owner to a card. - name: add_comment description: Post a comment on a card. - name: Cockpit and supervision identity: core tools: - name: get_pilotage_state description: Read workspace steering signals. - name: get_cockpit_dashboard description: Read the consolidated cockpit dashboard. - name: get_agent_status description: Read agent status. - name: list_pending_proposals description: List writes that autonomy policy converted into pending proposals. - name: Agent runtime identity: agent token tools: - name: stellary_init description: Bootstrap an agent loop; returns autonomy mode, rules, skills and the filtered tool list. - name: get_next_mission description: Claim the next queued mission. - name: wait_for_mission description: Long-poll for a queued mission. - name: complete_mission description: Complete a mission; updates mission state and posts a card comment. - name: fail_mission description: Fail a mission; updates mission state and posts a card comment. - name: Identity identity: any tools: - name: stellary_list_identities description: >- List the opaque identity references an OAuth connection authorized. With more than one identity, actingAs is required on every call. - name: Plugin-registered tools identity: agent/workspace context with decrypted plugin config note: >- Installed workspace plugins auto-register tools prefixed by plugin slug. Documented plugin families are GitHub, Slack, Email, Discord and X. The exact list is workspace-dependent, so only the documented examples are recorded here -- never assume the full set. tools: - name: github_list_repos - name: github_create_pr - name: slack_post_message - name: email_send_email - name: discord_post_message - name: x_create_draft governance: autonomy_modes: - mode: autonomous reads: execute writes: execute - mode: supervised reads: execute writes: safe tools execute; tools marked approval_required become persisted proposals - mode: approval reads: execute writes: every non-read tool becomes a persisted proposal note: >- Autonomy proposal checks apply to agent tokens only. A human acting through MCP acts as themselves and writes execute directly, still subject to permissions and scopes. clients_documented: - Claude Code (claude mcp add stellary --transport streamable-http https://api.stellary.co/mcp) - Cursor (mcpServers.stellary.url) - Claude Desktop - Gemini CLI (httpUrl, not url) - Codex / ChatGPT - LM Studio - AnythingLLM