Glama
The MCP server registry — discover, score, and inspect thousands of Model Context Protocol servers.
About
Glama is the directory the MCP ecosystem standardized around: a searchable registry of MCP servers where each listing gets an automated quality score (tool disambiguation, naming consistency, completeness), health probes with uptime history, and install instructions. The companion repo punkpeye/awesome-mcp-servers — 96,000 stars, updated daily — is effectively the registry's index, and half the READMEs in the MCP world now carry Glama score badges.
For agent builders it answers the practical questions: which server for this job, is it maintained, does it actually respond. Listings are claimable by maintainers via a glama.json in the repo, and the site has grown into connectors and a gateway alongside the directory. The caveats are directory-model ones: scores are automated heuristics, not audits — a high score is not a security review — and listing quality ultimately depends on maintainers keeping their servers healthy.
What it can do
- Searchable registry of MCP servers with automated quality scores
- Health probes and uptime history per server/connector
- Score badges for READMEs (disambiguation, naming, completeness)
- Claimable listings via `glama.json` maintainer verification
- Install instructions per server (npx/uvx commands, client JSON snippets)
- Connectors directory and gateway alongside the server registry
Setup
No install — it is a web directory. 1. Browse servers at glama.ai/mcp/servers or search by capability.
- Check a listing's score breakdown, health probes, and uptime before adopting it.
- Follow the listing's install instructions (most are one
npx/uvxcommand or a JSON snippet for your MCP client). - Maintainers: add a
glama.jsonto your repo's default branch to claim the listing and manage badges.
Prerequisites
An MCP-compatible client (Claude Code, Cursor, VS Code, or any MCP host) to use the servers you find; nothing to browse.
Use cases
Works with
Trust score — 9/10
Scored 9/10 (metrics tier, metrics-v1). Evidence: the directory repo carries 96,393 stars (GitHub API, used as an adoption proxy for the registry), MIT, pushed Oct 15 2026; the live registry shows scored listings with health probes and uptime; widespread badge adoption across MCP READMEs corroborates its de-facto-standard status. Strengths: the MCP discovery layer, automated scoring plus live health data, claimable listings. Docked for: (1) scores are heuristics, not security audits — verify before wiring a server into sensitive tools; (2) adoption metric is a proxy (repo stars, not registry accounts); (3) listing quality depends on maintainers. License: MIT (directory repo).
★ 96,393 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 9.0/10
- Maintenance · recency, releases, issue load
- 10.0/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · Adoption 9: proxy — starScore from 96,393 stars on punkpeye/awesome-mcp-servers (the registry's directory repo); badge adoption across MCP READMEs corroborates de-facto-standard status
- · Maintenance 10: repo pushed 2026-10-15 (recency 10); registry listings show live health probes as of Oct 15 2026
- · Openness 10: MIT directory repo + live registry verified
- · Manual 7: live registry shows scored listings with health probes and uptime; docked for scores being heuristics not security audits, proxy adoption metric, maintainer-dependent listing quality
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.