IMPORTANT
This server has been merged into commitshow-mcp and is no longer published.
commitshow-mcp carries the same two tools — search_services and fetch_docs, with
identical arguments — and adds audit_repo and project_status. Two entries under one
name, half of them the same, gave anyone browsing the registry no way to tell which to
install; there is one now.
{ "mcpServers": { "legitshow": { "command": "npx", "args": ["-y", "commitshow-mcp@latest"] } } }
The npm package and the registry entry io.github.commitshow/legitshow-search are both
marked deprecated. The source stays here as the record of what it was.
Legit.Show MCP server
Model Context Protocol server for Legit.Show.
Plug it into Claude Desktop, Cursor, Cline, Windsurf, or any MCP host, and your model
can search a directory of launched software by measured production-readiness
— not GitHub stars — without ever leaving the chat.
{
"mcpServers": {
"legitshow": {
"command": "npx",
"args": ["-y", "legitshow-mcp"]
}
}
}
That's the entire install. No API key. The server is a stdio shim over the public
https://legit.show/api/search REST surface — one source of truth for the data.
Why this is different
Most "find me a tool" flows rank by GitHub stars or popularity. Legit.Show carries a
deterministic 7-Frame production-readiness benchmark (Performance, Accessibility,
Security, Privacy, Reliability, Standards, Discoverability) measured from the public
surface — real Lighthouse, HTTP headers, live probes, no LLM in the scoring path. So
your agent can filter by whether something is actually production-ready:
"find production-ready, open-source MCP servers with Security ≥ 80"
| Tool | What it does |
|---|
search_services({ query?, category?, min_scores?, is_open_source?, limit? }) | Search launched software by name or filter. Returns each match with its per-frame scores and its legit.show/s/<slug> page. |
fetch_docs() | Pull the canonical Legit.Show docs (llms.txt) — what Legit.Show is, the 7-Frame benchmark, and how to cite a service. |
Boundary: per-frame scores + summary + the page only. There is no combined overall
total in the public data (it's a licensed product), so the model won't invent one.
Host wiring snippets
Claude Desktop — claude_desktop_config.json
{ "mcpServers": { "legitshow": { "command": "npx", "args": ["-y", "legitshow-mcp"] } } }
Cursor — .cursor/mcp.json
{ "mcpServers": { "legitshow": { "command": "npx", "args": ["-y", "legitshow-mcp"] } } }
Cline / Continue / Windsurf
Same three-line npx -y legitshow-mcp invocation in the host's MCP config.
No MCP host? Plain REST
GET https://legit.show/api/search?q=<name>
Filters: category, min_scores (e.g. {"security":80}), is_open_source, limit (≤50).
GET or POST. No key.
Self-hosting / dev overrides
| Env | Default | Used by |
|---|
LEGIT_SEARCH_API | https://legit.show/api/search | search_services |
LEGIT_DOCS_BASE | https://legit.show | fetch_docs (llms.txt) |
Links
- Legit.Show — the directory: every launched service, tested
- Methodology — how the 7-Frame benchmark is measured
- llms.txt — the canonical machine-readable doc
MIT · operated by Madeflo, Inc.