One tool, one job: you describe the idea while you are still describing it, and five sources answer before the repository exists. The score is less useful than what comes with it — the named projects, their star counts and the trend direction are what turn a hunch into build, pivot or drop.
A single-tool server. You describe an idea in plain language; it queries GitHub, Hacker News, npm, PyPI and Stack Overflow in parallel and comes back with a score, the projects already occupying that space, and where the interest is heading.
- A reality signal from 0 to 100, with a duplicate likelihood in words next to it
- A trend reading — accelerating, stable or declining — and a market momentum sub-score, so a crowded space and a growing one are told apart
- The evidence behind the score, source by source: repository and star counts on GitHub, mentions on Hacker News, package counts on npm and PyPI, question counts on Stack Overflow
- The closest existing projects by name, URL, stars and description — the competitor you would have found in week three
- Pivot hints pointing at where the leading project leaves room
- Two depths on the same call: quick reads GitHub and Hacker News, deep goes to all the sources
Python with uv, and nothing to configure — no account, no key. GITHUB_TOKEN is worth setting anyway: GitHub's unauthenticated rate limit is what a run of deep checks hits first. The project is in maintenance mode — it works and stays free, but no new features are planned. MIT licensed.
One command — uvx idea-reality-mcp