Provides integration with NVIDIA's OpenAI-compatible endpoint (e.g., z-ai/glm-5.2) for generating product planning outputs, supporting structured JSON output.
Integrates with any OpenAI-compatible Chat Completions API to generate structured product planning packages including PRDs, requirements, user flows, wireframes, data schemas, API contracts, and MCP handoff resources.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@ai-product-planner plan a habit tracker app with streaks and reminders"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
AI Product Planner turns a product idea into an implementation-ready planning package: PRD, requirements, user flow, wireframes, data schema, API contracts, SDK boundaries, and an MCP handoff for coding agents.
The generation pipeline uses a supervisor stage followed by dependency-aware parallel workers. It calls any OpenAI-compatible Chat Completions API and keeps generated sessions on the local filesystem.
Node.js 22 or newer
An OpenAI-compatible LLM API key and model with structured JSON output support
Related MCP server: Gigaspec
npm install
cp .env.example .env
# Set PLANNER_LLM_API_KEY and, if needed, the base URL and model.
npm run build
npm startOpen http://127.0.0.1:8792.
For frontend and backend development in separate terminals:
npm run dev:server
npm run devThe default example targets NVIDIA's OpenAI-compatible endpoint with z-ai/glm-5.2. Other providers work when they implement POST /chat/completions with messages, response_format, and choices[0].message.content.
Set PLANNER_LLM_STRUCTURED_OUTPUT=false only when the selected provider does not support response_format: { "type": "json_object" }. The planner still validates the returned JSON and fails explicitly if the contract is invalid.
The default server binds only to 127.0.0.1 and allows unauthenticated local use. It refuses to start with PLANNER_AUTH_MODE=none on a non-loopback host.
For deployment behind an access proxy, use header mode:
HOST=0.0.0.0
PLANNER_AUTH_MODE=header
PLANNER_AUTH_HEADER=x-authenticated-user
PLANNER_AUTH_HEADER_VALUE=expected-proxy-valueYour proxy must strip incoming copies of that header and inject the trusted value after authentication. MCP clients can instead use PLANNER_MCP_BEARER_TOKEN.
The JSON-RPC MCP endpoint is /mcp. It exposes planning sessions, generated contracts, active implementation goals, and run handoff resources.
curl http://127.0.0.1:8792/mcp \
-H 'content-type: application/json' \
-d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{}}'npm test
npm run lint
npm run typecheck
npm run buildSee SERVER_START_HERE.md, SERVER_OPERATIONS.md, and APP_DATA_SCHEMA.md for integration details.
MIT
This server cannot be deployed
Turn your app idea into IA, wireframes, PRD, style guides, and dev specs for coding agents.
Turn PRDs and product ideas into structured specs so coding agents build your intent, not theirs.
Turns vague automation requests into tool stacks, prompts, QA checks, and human boundaries.
AI workflow/MCP implementation package planner.
