About MCP Server Finder

• Server listings come from a curated seed set, the official MCP registry (registry.modelcontextprotocol.io), and community submissions that are reviewed by a human before publication.
• Stars, forks, open issues, license, and commit activity are pulled from the public GitHub API and snapshotted once a day.
• Every server page lists the source and confidence of each field, so you can check where a value came from.

Each server gets popularity, freshness, growth (7-day star change), and maintenance scores from 0 to 100. The overall score weights them 35% / 25% / 25% / 15%. Scores are recomputed after every daily data refresh; nothing is ranked by hand or paid placement.

Benchmarks check installability, startup, capability discovery (the tools, resources, and prompts a server exposes), documentation, configuration, and safety. Servers are only started in an isolated environment without real credentials. Checks that cannot run safely are recorded as skipped — never as a pass.

Risk profiles are inferred from each server's README and metadata (API tokens, filesystem access, external writes, database and network access). When the evidence is insufficient the level is shown as Unknown.

MCP Server Finder is not affiliated with Anthropic or the Model Context Protocol project. The site is supported by advertising, which does not affect rankings or scores.

Found an error or want a server listed? See the contact page.