Free MCP check
MCP check registry
Open-source MCP servers listed by their maintainers, graded from their source and kept current by mcp-check in CI.
How it works
Step 1: Submit
Open the submission form in the registry repository with your server’s GitHub link. A server is listed only when its maintainers submit it.Step 2: Get graded
mcp-check reads the tool definitions from the source at one commit and grades five areas, A to F, with advice on the write surface. It never runs or calls the server. The grades and the fix list come back in your issue.Step 3: Add to CI
One click proposes a workflow that runs mcp-check on every push. With it on your default branch, the listing is marked CI-verified and stays current.
Listed servers
3 servers listed.
| Server | Status | Grades | Findings |
|---|---|---|---|
| Kronicle Liteisokron-ai/kronicle-liteLocal-first MCP server: a hash-chained document store, signed retrieval receipts and an orient protocol for stale agent beliefs. | Not in CI | Lowest grade: AMCP specification 2026-07-28: AAnthropic connector directory: AOpenAI plugin directory: APoisoning and hidden text: ASchema and description quality: A | 0 high · 2 medium · 2 low |
| SEA Liteisokron-ai/sea-liteFenced work queue for parallel coding agents: leased claims, path-scoped file locks, typed links and hash-chained history. | Not in CI | Lowest grade: BMCP specification 2026-07-28: AAnthropic connector directory: AOpenAI plugin directory: BPoisoning and hidden text: ASchema and description quality: A | 4 high · 5 medium · 40 low |
| IsoKron MCP checkisokron-ai/mcp-checkGrades an MCP server's tool list against the MCP spec and the Claude and ChatGPT directory rules. CLI, Action and MCP server. | Not in CI | Lowest grade: CMCP specification 2026-07-28: CAnthropic connector directory: AOpenAI plugin directory: APoisoning and hidden text: ASchema and description quality: A | 0 high · 6 medium · 5 low |
What a listing shows
- Five grades from the checker’s rubric: MCP specification 2026-07-28, Anthropic connector directory, OpenAI plugin directory, Poisoning and hidden text, Schema and description quality. The write surface is advice, not graded, since rubric 2.2.0. Each links to the checks behind it.
- The findings by severity, the commit graded, the checker and rubric versions, and the scan date.
- CI-verified when a workflow on the default branch uses
isokron-ai/mcp-check. - Owner-verified when the default branch has
.github/mcp-check/registry.jsonnaming the listing. No sign-in is needed. - A README badge that links to the listing.
Grades come from reading the source statically; nothing is run. They say what the tool definitions show, not how the server behaves when it writes: that is what the MCP Write-Safety Requirements test. To question a grade or have a listing removed, see Disputes or open an issue in the registry repository.
Add it to your CI
Enter your repository and GitHub opens with the workflow filled in. You review it and commit it yourself. It reads the tool definitions from your source on every push and pull request, with read-only permissions.
name: MCP check
on:
push:
pull_request:
permissions:
contents: read
jobs:
mcp-check:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
with:
persist-credentials: false
- uses: isokron-ai/mcp-check@v1
Request an audit or a build
Tell us which server and what you need. We reply to confirm the scope, the price and the dates.