MCP Integration

Tools & resources

The exact, source-verified tool and resource lists for both the local stdio and remote HTTP MCP servers — and how they currently differ.

These two servers expose different tool sets. The local server (packages/cli/src/mcp/server.ts, started by intent mcp) and the remote server (src/lib/mcp-server.ts, served at /api/mcp) are two separate implementations that have drifted apart — the remote server has a newer, larger tool set (including batch operations and a start_story/complete_story pair) that the local server doesn't have yet. Build against whichever one you're actually pointed at; don't assume tool names are interchangeable between them.

Local server (`intent mcp`) — 13 tools

Tools are gated by --mode (core, standard, or all) — each mode includes everything from the tiers below it.

Core

Tool	Description
`list_stories`	List all stories from the story map with status, priority, effort, and acceptance criteria
`mark_done`	Mark a story as done. Blocks if a `must` acceptance criterion is failing or unreported, unless `force: true`
`report_criteria`	Report pass/fail/blocked results for one or more acceptance criteria on a story
`raise_blocker`	Raise a blocker on a story after repeated failed fix attempts
`mark_in_progress`	Mark a story as in progress
`get_status`	Progress summary — total, done, in progress, todo counts
`add_note`	Add implementation notes to a story

Standard (adds)

Tool	Description
`create_story`	Create a new story — title, activity, priority, effort
`get_context`	Get the full `INTENT.md` context — personas, stories, acceptance criteria
`log_decision`	Log an architectural or product decision

All (adds)

Tool	Description
`analyze_complexity`	AI complexity scoring — 1–10 score, risks, suggested effort
`research`	AI research using your project context
`import_prd`	Parse a PRD or spec into structured stories

Resources (always available, regardless of mode)

Resource	URI	MIME type
Story Map	`intentdocs://project/storymap`	`text/markdown`
Stories	`intentdocs://project/stories`	`application/json`
Status	`intentdocs://project/status`	`application/json`
Decisions	`intentdocs://project/decisions`	`application/json`
Data Model	`intentdocs://project/data-model`	`application/json`
Build Order	`intentdocs://project/build-order`	`application/json`

Resources are read-only and auto-discoverable — an agent can read intentdocs://project/data-model directly without calling a tool first.

One more behavior worth knowing: if a story's cancelled_at is set (via intent stop), the local server refuses every tool call on it except list_stories and get_status — a cancellation guard that stops an agent mid-flight from continuing to work a story a human just pulled the plug on.

Remote server (`/api/mcp`) — 20 tools, always all-on

No mode gating — every tool below is always available to an authenticated caller. There are no resources on this server; everything is a tool call.

Tool	Description
`list_projects`	List all your IntentDocs projects
`get_context`	Get the full `INTENT.md` context for a project's story map
`list_stories`	List all stories in a project's story map with status, criteria, and build order
`update_stories`	Update story fields — status, notes, implementation details, or user-story text (`as_a`/`i_want`/`so_that`)
`create_stories`	Create new user stories in a project's story map
`report_criteria`	Report acceptance criteria test results for a story
`add_criteria`	Append acceptance criteria to a story — returns new criterion IDs for use with `report_criteria`
`replace_criteria`	Replace all acceptance criteria on a story — returns new criterion IDs
`check_release_readiness`	Pre-flight check before building a release — scans for missing ACs, missing implementation plans, missing test criteria
`start_story`	Begin work on one story — sets `in_progress`, returns its acceptance criteria, implementation plan, and context
`complete_story`	Finish a story: mark done, record what was built, get the next story. Enforces the must-criteria gate unless `force: true`
`link_pr`	Link a GitHub pull request to a story — the PR number appears on the story card in the UI
`get_status`	Progress summary for a project
`raise_blocker`	Raise a blocker on a story after repeated failed attempts
`add_note`	Add implementation notes to a story
`log_decision`	Log an architectural or product decision
`analyze_complexity`	AI complexity scoring for stories
`research`	AI research using your project context
`import_prd`	Parse a PRD or spec into stories
`get_build_order`	Get the recommended build order for stories in a project

update_stories/create_stories are batch operations (arrays of stories), unlike the local server's singular create_story. start_story/complete_story replace the local server's mark_in_progress/mark_done with a workflow that also hands back the next story to work on.