Why another list
The awesome-* convention has served open source well, but it has a known failure mode: the list is a file in a repository, and files do not notice when the projects they link to are abandoned. Popular awesome lists routinely carry entries that stopped being maintained a year ago, because removing something requires a human to notice and open a pull request.
This list is generated instead. It comes from the same dataset that powers the rest of this site, refreshed daily, so a project that stops getting commits or gets archived simply stops appearing here. Nobody has to remember to remove it.
The trade-off is that a generated list cannot capture taste. It cannot tell you that one server has a nicer API than another, or that a particular maintainer is responsive. For that, read the detail pages and the repositories themselves — this list is a filter, not a verdict.
The list
Actively-maintained servers with at least 1,000 stars that are verified in the official MCP registry, ordered by TrustScore. Refreshed daily.
| # | Server | TrustScore | Stars | Clients |
|---|---|---|---|---|
| 1 | io.github.D4Vinci/ScraplingScrapling MCP Server | 93 | 73,830 | 5 |
| 2 | io.github.ChromeDevTools/chrome-devtools-mcpChrome DevTools MCP | 93 | 49,416 | 5 |
| 3 | io.github.modelscope/funasr-mcpFunASR | 92 | 19,893 | 5 |
| 4 | com.mock-server/mockserverMock, record/replay, verify and chaos-test any HTTP, REST, gRPC or LLM dependenc | 89 | 4,942 | 4 |
| 5 | io.github.netdata/mcp-serverNetdata | 87 | 80,154 | 4 |
| 6 | io.github.microsoft/playwright-mcpPlaywright Tools for MCP | 87 | 36,128 | 5 |
| 7 | io.github.homeassistant-ai/ha-mcpComprehensive Model Context Protocol server for managing Home Assistant through | 87 | 4,381 | 5 |
| 8 | io.github.bytebase/dbhubDBHub | 87 | 3,365 | 5 |
| 9 | io.github.ihor-sokoliuk/mcp-searxngSearXNG Search | 87 | 1,115 | 5 |
| 10 | io.github.usekaneo/kaneoOfficial MCP server for Kaneo: manage tasks, projects, and labels from Claude an | 86 | 8,047 | 5 |
| 11 | io.github.Ataraxy-Labs/semEntity-level code intelligence: semantic diff, impact analysis, blame, and conte | 86 | 3,302 | 5 |
| 12 | io.github.DeusData/codebase-memory-mcpCodebase Memory | 85 | 39,581 | 5 |
12 of 2861 actively-maintained servers in these categories cleared the 1000-star adoption floor, ordered by TrustScore. Regenerated daily from public GitHub, npm and registry signals — see our editorial policy for the full weighting. ✓ means we installed it in a sandbox and confirmed it starts.
How to read it
TrustScore combines five dimensions of public data: maintenance activity, adoption, usability, repository health and community signals. The weighting is published in full on our editorial policy page. It measures whether a project is alive and used — not whether it is secure, and not whether it fits your particular problem.
Star counts are included alongside the score because the two answer different questions. A high score with modest stars usually means a well-run but niche project; high stars with a middling score often means something popular that has slowed down. Neither is inherently better.
The clients column shows how many MCP clients the server can connect to, derived from the package type and transport it declares. A check mark means we installed it in a sandbox and confirmed it starts — measured, not inferred.
What is filtered out, and why
Projects under 1,000 stars. This list is deliberately a high bar; our full catalog covers everything else and lets you filter by category.
Anything not verified in the official registry. For a broad cross-category list this filter does real work — it removes entries whose metadata nobody has checked.
One entry that is registry-verified but declares a repository it does not own: its manifest points at an unrelated project with 88,000 stars, which would have placed it near the top on borrowed adoption. The error is upstream in the registry and we cannot fix it there, but we are not going to republish someone else's stars as theirs.