Intent Compiler
Build structured intent specs through Socratic conversation in your terminal. Zero config. No signup. Just add the MCP server and start talking.
Works with Claude Code, Cursor, and Codex today.
How it works
Claude becomes your product thinking partner. It challenges vague ideas until the spec is clear enough to build.
1. Describe the problem
Paste a support ticket, describe a user complaint, or just say what's broken. Claude starts by challenging your framing.
2. Socratic conversation
Claude pushes back on vague language, asks one pointed question at a time, and proposes precise spec fields as clarity emerges.
3. Save & export
When the spec is ready, Claude writes intent.md to your project and optionally exports as AGENTS.md, CLAUDE.md, or .cursorrules.
What you get
A structured spec that AI coding agents can actually execute. Not a vague description — your product judgment made explicit: the objective, testable outcomes, and the edge cases that usually stay in someone's head.
- intent.md
The structured spec file. YAML frontmatter + markdown sections. Readable by humans, parseable by tools.
- AGENTS.md
The native instruction file for OpenAI Codex and modern Cursor. Same intent context, written where those agents look for it.
- CLAUDE.md
Appends a Pathmode section to your project's CLAUDE.md. Every Claude Code conversation sees the intent.
- .cursorrules
Agent-directive format for legacy Cursor setups. Newer Cursor versions read AGENTS.md instead.
---
id: "intent_17291..."
version: 1
status: "draft"
---
# Fix Checkout Payment Timeout
## Objective
Reduce cart abandonment caused by payment step
timeouts exceeding 3 seconds, where users see
a spinner with no status feedback or recovery path.
## Outcomes
- [ ] Payment completes in under 3 seconds for
95th percentile of transactions
- [ ] Users see real-time status during processing
- [ ] Failed payments show actionable error with
retry button
## Edge Cases
- **Network timeout during payment**: Show retry
button, do not double-charge
- **Payment provider returns unknown status**:
Hold order, notify user within 30 secondsGet started in 30 seconds
Add the Pathmode MCP server to your Claude Code settings. No API key needed.
Add to your Claude Code settings
{
"mcpServers": {
"pathmode": {
"command": "npx",
"args": ["@pathmode/mcp-server"]
}
}
}Start a conversation
> Help me write an intent spec for improving our search resultsClaude picks up the compile-intent prompt and becomes your Socratic product thinking partner.
Save and export
When the spec is ready, Claude automatically calls intent_save to write intent.md to your project root. Want agent-ready output? Ask for AGENTS.md, CLAUDE.md, or .cursorrules export too.
Ships with a Claude Code skill pack
Seven skills install alongside the MCP server and auto-trigger on what you ask — no slash commands to memorize. Build, stress-test, verify, and hand off specs without leaving the terminal.
compile-intentBuild a structured spec for what to ship
grill-intentStress-test an existing spec for weaknesses
verify-intentDesign the executable feedback loop for a spec
review-against-intentCheck code changes against intent outcomes
split-intent-to-issuesBreak a spec into Linear, Jira, or GitHub issues
handoff-intentCapture decisions at the end of a session
setup-pathmode-workflowFirst-time setup: test commands, tracker, conventions
npx @pathmode/mcp-server install-skills
Installs to .claude/skills/ in your project — add --global for user-wide. Full docs
From REPL to Runtime
The Intent Compiler gives you structured specs in your terminal. When you need dependency tracking, team collaboration, and governance — connect to the full platform.
Intent Compiler
Free, zero-config
- Socratic spec building in terminal
- Export to intent.md, AGENTS.md, CLAUDE.md, .cursorrules
- 7 auto-triggering Claude Code skills
- Works offline, no account needed
- Single-intent, single-developer
Pathmode Platform
For builders shipping with AI
- Everything in Compiler, plus:
- Intent dependency graphs & critical path
- Evidence board — link specs to user feedback
- Team collaboration & governance
- Auto-sync to Linear and Jira