Supaspec stores versioned implementation specs that AI agents read and write via MCP. Write once. Every agent stays in sync.
Your spec is the single source of truth. Agents connect via MCP.
Write an overview. AI generates the full spec. Multiple agents review and refine — architecture, security, market research — all tracked.
Point a coding agent at your spec. It reads every section and builds. Status tracking shows progress in real-time.
Write feedback as spec sections. Agents read it, fix it, mark it done. No context lost.
Claude Code, Cursor, Codex, Lovable, Replit — they all work great alone. Supaspec gives them a shared spec to read and write.
Create sections, set statuses, write in markdown.
One command to connect via MCP.
Track every change. Accept or reject proposals.
Every save is a commit with message, author, and timestamp. Diff any two versions.
Write directly or suggest changes for review. Your choice per update.
See which AI made each change. Color-coded per agent across the editor.
Works with Claude Code, Cursor, Codex, Lovable, Replit out of the box.
Search across all sections in a project. Find anything in your spec instantly.
Multiple agents propose changes to the same section. Cherry-pick the best from each.
Every change carries a message, description, and author. Like git log for your spec.
See who's editing what, when. Agent activity dashboard with revision history.
Add Supaspec as an MCP server to your AI coding tool. One command, and your agents can read and write your spec.
claude mcp add --transport http supaspec https://supaspec.dev/api/mcpforever
per user