MCP Radar

Why trust this leaderboard?

Because the scoring rules, data sources and decision logic are all written on this page, and every server page notes its data sources and fetch time. No black box — you can re-check any score with the same public data.

👤Maintainer

MCP Radar is maintained by an independent developer who has long tracked the AI tooling ecosystem. The reason for building this is simple: existing MCP directories only tell you "this tool exists", nobody tells you "whether it's still alive" — you install an abandoned server and debug for hours before realizing the problem isn't yours. That judgment can be fully automated, so we did it, and made it fully public.

— MCP Radar maintainer (brand byline, real name can be swapped in later)

Scoring methodology (fully public)

TrustScore is out of 100, weighted across five dimensions, all signals from public APIs:

DimensionWeightSignal fields
Maintenance30%Days since last commit, commits in 90 days, average issue response, whether archived
Adoption25%GitHub stars and trend (weighted above npm downloads), npm weekly download trend, release frequency
Usability20%In official registry, server.json spec, whether a runnable entry can be resolved
Health15%Open issue/PR backlog ratio, whether license is clear
Community10%Contributor count, fork activity, awesome-list inclusions

Lifecycle rules

🟢 active

Normally maintained

🟡 dying

Last commit > 180 days AND no issue response AND no new release

⚰️ dead

Repo archived == true

unverifiable

Remote-only (no open-source repo / no package), unverifiable

"Has an auditable repo" is a usability bonus — a remote-only service can't have its behavior verified, which is itself a risk signal.

Data sources and limits

SourceFields providedKnown limits
MCP official registryManifest, repo url, package name, official status, publish/update timeNo stars / activity / downloads
GitHub APILast commit, commits in 90 days, issue response, archived, stars, forks, open issues, license, contributors
npm registryWeekly downloads, version release timelineUnderestimates servers installed directly from GitHub

Update frequency: health data is scanned incrementally daily, with a full diff weekly (the radar page's data source).

We flag our boundaries: remote-only servers have no repo signals and are always marked unverifiable; npm downloads underestimate servers installed directly from GitHub, so stars are weighted higher; fetching has lag, so a project that just resumed maintenance may still be briefly marked dying — which is exactly why we keep a correction channel open.

Conflict-of-interest disclosure

This site accepts sponsored detail-page placements (clearly labeled "Sponsored"), but rankings, scores and categories never accept bidding of any kind. Sponsorship buys placement, not a score. The moment we accept ranking bids, the credibility of our proprietary data collapses — that's the foundation of this site, not for sale.

See sponsorship rules →

Data correction / death-call appeal

Maintainer thinks a server's score or call is wrong (e.g. maintenance resumed, repo moved)? Send the repo URL and details; we review manually and update within 48 hours: corrections@mcpradars.com