analyze_intent_graph
Analyze the intent dependency graph for risks and strategic insights. Returns critical path, cycles, bottlenecks, orphans, status mismatches, and stalled intents.
LABSCO SUMMARY
A spec an agent can be graded against, and correct when the code proves it wrong.
Two tools carry the idea: verify_implementation scores what was built against each outcome and constraint separately, and record_implementation_finding sends the correction back into the spec instead of leaving it in a commit message. The specs stay as markdown in your repo, so nothing here holds your requirements hostage.
What it is
An MCP server for Pathmode's intent specs. Specs live as plain markdown you own — intent.md in the repo — and the server gives an agent the structured version: objectives, outcomes, constraints, edge cases, the dependency graph between intents, and the evidence behind why each one exists.
What you get
- get_current_intent, get_intent, list_intents and search_intents — the active intent with a summary of the evidence supporting it, one by ID, all of them with status, or a keyword search across goals, objectives, outcomes and constraints
- get_intent_relations and analyze_intent_graph — what an intent depends on, enables or blocks, and an analysis of the whole graph: critical path, cycles, bottlenecks, orphans, status mismatches and stalled work
- get_agent_prompt — the full execution prompt for one intent, with its objective, outcomes, constraints, edge cases and verification steps
- get_workspace and get_constitution — the vision, non-negotiables and architecture principles, and the constitution rules every implementation must respect
- create_intent, update_intent and update_intent_status — write a new spec, change its content, or mark it shipped or verified
- verify_implementation — grade what you built against the spec, item by item across outcomes, constraints, constitution rules and edge cases, with a pass or fail and reasoning for each
- log_implementation_note and record_implementation_finding — record why you chose an approach, and correct the spec when building it proved an assumption wrong
- query_evidence, create_evidence and link_evidence — friction points, user quotes, observations, metrics and feature requests, searchable and attachable to the intents they justify
- export_context, sync_context, intent_save and intent_export — write the workspace context into CLAUDE.md or AGENTS.md idempotently, save a spec to intent.md, or export it as .cursorrules or an outcomes rubric
Requirements
Nothing to sign up for. Add it to .mcp.json and the client runs npx @pathmode/mcp-server; specs live in intent.md in your repository. Set PATHMODE_API_KEY only if you want intent and evidence synced across a team.
Setup effort
One command — npx @pathmode/mcp-server
USE CASES
01
Starting work on a feature
Read the active intent, then get the full execution prompt with its constraints and verification steps.
02
Checking you actually built it
Grade the implementation against the spec item by item and read the reasoning behind each failure.
03
Fixing a spec the code disproved
Write the correction back into the intent the moment you find the assumption was wrong.
04
Deciding what to do next
Analyse the dependency graph for the critical path, bottlenecks and stalled intents, and read the evidence behind the candidates.
05
Giving the whole repo the same context
Sync the workspace context into CLAUDE.md or AGENTS.md so every agent starts from it.
22 tools
Counted once with only startup setup. Other versions, modes or optional credentials may expose a different list.
create_evidence
Create a new evidence item (e.g., a discovered bug, user feedback quote, behavioral observation, or feature request). Evidence can later be linked to intents to support prioritization.
create_intent
Create a new intent spec in the workspace. Requires at minimum a title, objective, and productId. Returns the created intent with its ID. Use list_intents first to see existing intents and avoid duplicates.
export_context
Export workspace context as a formatted file. Use "claude-md" for CLAUDE.md (full workspace context), "agents-md" for the same workspace context written for AGENTS.md (the native instruction file for OpenAI Codex and modern Cursor), "cursorrules" for Cursor AI rules, "intent-md" for a single intent specification file, or "outcome-rubric" for a Claude Managed Agents Outcomes rubric (writer description + evidence-forcing grader rubric) derived from the resolved intent, its implementation context, and the workspace constitution. claude-md and agents-md produce identical, agent-agnostic content (the difference is only the target filename). For cursorrules/intent-md/outcome-rubric, product context is always derived from the resolved intent. For claude-md/agents-md, pass productId to select a specific product, otherwise the first active product is used.
get_agent_prompt
Get a formatted execution prompt for a specific intent. This is the full structured prompt including objective, outcomes, constraints, edge cases, and verification steps.
get_constitution
Get the workspace constitution rules. These are mandatory constraints that all implementations must respect.
get_current_intent
Get the currently active intent, preferring approved intents with real objective/outcome content over empty stubs. Returns the full IntentSpec plus an evidenceSummary (how much user evidence backs it). Evidence is referenced by ID; pass include_evidence=true to inline the evidence content, or call get_agent_prompt for the full evidence-backed execution prompt.
get_intent
Get a single intent by ID with full details including objective, outcomes, constraints, edge cases, and relations. Evidence is referenced by ID; call get_agent_prompt for the full evidence-backed execution prompt.
get_intent_relations
Get the dependency graph for a specific intent. Shows what it depends on, enables, or blocks.
get_workspace
Get workspace details including strategy (vision, non-negotiables, architecture principles), active products, and constitution rules.
intent_export
Export an intent spec as .cursorrules, a CLAUDE.md or AGENTS.md section, or a Claude Managed Agents Outcomes rubric for AI agent consumption. Use agents-md for Codex, Cursor, and other AGENTS.md-aware agents.
intent_save
Save an intent spec to intent.md in the project root. Called after building a spec through conversation.
link_evidence
Link or unlink evidence items to/from an intent. Linking evidence to intents establishes traceability between user problems and planned solutions.
list_intents
List all intents in the workspace. Returns an array of IntentSpecs with their status, objectives, and metadata.
log_implementation_note
Record a technical decision or implementation note for an intent. Use this to document why you chose a specific approach.
query_evidence
Search evidence items (friction points, user quotes, observations, metrics, feature requests) by product, type, severity, or text. Returns matching evidence with IDs that can be linked to intents.
record_implementation_finding
Write a correction back to an intent when BUILDING it revealed the spec was wrong. Use this the moment you discover the spec assumed something the implementation contradicts — instead of only fixing it in your head or this chat, record it so the NEXT agent or person inherits the correction. This is the inverse of a decision: it closes the refinement loop so the same stale premise is not rebuilt. Open findings ride into future agent prompts until a human reconciles them into the spec.
search_intents
Search intents by keyword. Searches across user goals, objectives, outcomes, and constraints.
sync_context
Write this workspace's canonical Pathmode context into the repo's CLAUDE.md (or AGENTS.md), idempotently — the round-trip. Pulls the latest from Pathmode and replaces the PATHMODE-marked section in place, so the repo's agent instructions never drift from the source of truth. Unlike export_context (which returns the text for you to read), this writes the file; re-running when nothing changed reports "no drift". Run it after the canonical context changes in Pathmode.
update_intent
Update an existing intent's content. Provide only the fields you want to change. Does NOT change intent status (use update_intent_status for that).
update_intent_status
Update the status of an intent. Use this to mark an intent as shipped after implementation, or verified after testing. For shipped/verified transitions, the response includes a verification checklist of outcomes, constitution rules, and health metrics that should be confirmed.
verify_implementation
AI-grade your implementation against the intent spec. Checks each outcome, constraint, constitution rule, and edge case. Returns pass/fail per item with reasoning and an overall score. Also logs the result as an implementation note.
BEFORE YOU INSTALL
What to know first
Tools you get
22 tools when we connected.
Requires
npx on PATH
CONTEXT COST
Moderate Standing cost
Rated by the size of the 22 tool definitions that ride along on every turn, not by how many there are
INSTALLATION
Connect your client
Add the block to your client, or run the command directly.
Server setup
Configuration file
JSON
{
"mcpServers": {
"pathmode": {
"command": "npx",
"args": [
"@pathmode/mcp-server"
]
}
}
}npx
TERMINAL
npx @pathmode/mcp-server
