MCP servers that ship as code — a repository or a package you install and run locally (usually over stdio). They have no hosted endpoint, so we cannot probe or grade them: instead each one shows its repository, install command, license, stars and when it was last updated.
Collected from GitHub repositories and npm, PyPI, Cargo, Go, Maven, NuGet and RubyGems packages tagged as MCP servers, the official MCP Registry, Docker's MCP catalog, ModelScope, Baidu MCP World, Cline and other public directories; one entry per repository. A server that also offers a hosted endpoint is listed under Remote instead. These entries are not counted in our reachability, verification or readiness statistics.
-
sidekernelgithub.com/sidekernel-ai/sidekernelupdated this year
An MRI for reasoning that reveals blind spots in AI-generated solutions. Multi-perspective architectural reasoning via MCP sidecar.
npm MIT -
MCP server that mines git history to surface code hotspots, file coupling, and review blind spots.
npm MIT -
🧠 Local-first AI memory OS — a shared knowledge graph for Claude, Cursor, Cline and any MCP client. Not just storage: it spots your blind spots and challenges your thinking. 25 MCP tools, fully offline.
★ 6 -
Citation-verified RAG over EIPs: withholds an answer when a citation cannot be checked, and documents the verifier's own blind spots.
MIT -
woodpecker-mcpgithub.com/sspcodeflix/woodpecker-mcpupdated this year
MCP server that gives an LLM agent (e.g. HolmesGPT) a materialized, queryable service dependency graph in a graph database - deterministic root-cause (deepest failing service), blast radius, and observability blind-spot detection.
pypi Apache-2.0 -
Catch the known pitfalls before your agent starts — 121 evidence-backed pitfall cards across seven domains, three-zone honest reports (verified / unverified / declared blind spots) plus executable verifiers. Skill + MCP engine + hooks.
npm MIT -
reasonforge-publicgithub.com/euuuuuuan/reasonforge-publicupdated this year
A deterministic, read-only reasoning harness that a coding agent can call over 8 MCP tools: it forces a scope lock, a plan whose first move is the cheapest way to be proven wrong, a blind-spot sweep with line-level evidence, and a verdict gate that refuses a verified-fact claim without proof of execution. Read-only is enforced by a test.
Apache-2.0