Most stack advice fails because it answers a question you have not framed yet; this one uses elicitation to build the requirements before it recommends anything, and stays on managed platforms so the recommendation ends in a deployment rather than an architecture debate. Recommendations come from rules over your answers, so treat the explanation as the useful part.
A tech-stack planning MCP server. Rather than returning a framework comparison, it asks about the project — what you are building, who uses it, how many of them, budget, and how technical you are — and turns the answers into a concrete stack with deployment steps.
- An interactive planning session that gathers requirements through MCP's elicitation spec, in plain language — `start_project_planning`
- A stack recommendation built from those requirements, usable with or without the interview — `recommend_stack`
- The reasoning behind the picks at a `basic` or `detailed` level — `explain_recommendation`
- Step-by-step deployment instructions for the platform you land on — `get_deployment_guide`
- A bias toward managed platforms — Vercel, Netlify, Supabase, PlanetScale, Firebase, Render — so the answer is something a small team can actually run
Python with `uv pip install fastmcp`, then `python run_server.py`. The documented client entry points at `run_server.py` by absolute path, so the checkout lives on the machine running the client. The project is `vibe-coder-stack-planner` 0.1.0.