A directory can help you find a name. MCP Queen is built for the next decision: whether the available public evidence is strong enough to spend time integrating, testing, or approving that server.
| Question | Evidence MCP Queen publishes | What it does not claim |
|---|---|---|
| Does it answer? | Dated reachability, protocol handshake, latency, and capability-discovery observations. | Future uptime or service-level guarantees. |
| What can it do? | Tool names, descriptions, and schemas observed from tools/list, searchable by capability. | That every tool behaves as advertised or is safe to call. |
| What is the trust evidence? | Separate security/access, data-integrity, citation, claim-verification, response-benchmark, and reviewed field dimensions. | A composite safety, compliance, or certification score. |
| Is the source location current? | Registry identity, mutable repository mapping, operational endpoint, and dated repository-mobility findings shown separately. | That a repository state proves endpoint state, or the reverse. |
| What happened in real use? | Quarantined reports reviewed for specificity and published as qualitative field evidence. | Votes, popularity rankings, or automatic grade changes. |
1. Search observed capabilities
Find the tools a live probe actually saw, not only category tags or a README claim. This is useful when the query is a task or data type rather than a product name.
2. Inspect the points behind the grade
Operational grades are deterministic summaries of reachability, protocol behavior, tool metadata, latency, and provenance. Every grade keeps its evidence trail and probe date.
3. Refuse the blanket trust score
Security, data integrity, citations, advertised claims, response behavior, and field experience remain independent. Missing evidence stays unaudited instead of becoming false confidence.
4. Add transparent field evidence
Real-use reports are quarantined and reviewed for specificity before publication. They complement repeatable probes without becoming stars, votes, or an invisible moderation signal.
5. Track identity separately from location
A Registry name, GitHub repository, and remote endpoint are different objects. MCP Queen's repository-mobility work exposes moved, archived, or unavailable mappings without rewriting operational history.
6. Serve humans and agents natively
The same evidence is available as readable pages, JSON APIs, llms.txt, and seven MCP tools. An agent can discover and inspect evidence without scraping the website.
A server can be reachable but over-privileged, well documented but stale, citation-rich but operationally unavailable, or secure at its public handshake while unaudited behind authentication. One number hides those distinctions. MCP Queen keeps the evidence dimensions separate so a human or agent can apply the policy appropriate to the task.
Security and verification roadmap rule
Broader probes can be added only as dated, method-labeled observations. Until a probe and its evidence are live, that dimension remains unaudited. New probes do not retroactively turn an operational grade into a security certification.
Researcher or operator
Search by task, compare candidates, and inspect caveats before spending time on setup.
Agent builder
Add one read-only evidence endpoint and keep discovery inside the agent workflow.
Security or platform team
Search dated security/access observations, retain unaudited gaps, and use the evidence as one input to allowlist or procurement review.
MCP maintainer
Check the current evidence page, preserve identity through repository moves, and install a live operational badge.
Request a scoped public-evidence review covering Registry identity, repository mappings, endpoint health, and maintainer rollout. The review does not create unsupported security or adoption claims.
What makes MCP Queen different from an MCP directory?
MCP Queen combines official Registry identity with observed tool catalogs, bounded operational probes, separate Trust Receipt dimensions, repository-mobility evidence, reviewed real-use field reports, and machine-readable MCP and JSON access. It does not collapse those signals into a blanket safety claim.
Is MCP Queen an MCP security scanner?
It publishes dated security and access-boundary observations where a defined probe has evidence, but it is not a security certification or a substitute for source review, deployment controls, least privilege, runtime monitoring, or an authorized security test.
How does MCP Queen handle stale or moved MCP repositories?
The namespaced Registry name remains the evidence key while the repository mapping and operational endpoint are shown separately. Dated mobility audits can identify renamed, transferred, archived, or unavailable public repository mappings without treating repository state as endpoint state.
Are MCP Queen field reports ratings or votes?
No. Reports must describe actual use, enter a quarantine queue, and are reviewed for specificity before publication. Accepted reports remain qualitative evidence and never change an operational grade automatically.
Can another AI agent query MCP Queen?
Yes. MCP Queen is itself a public Streamable HTTP MCP server and also publishes JSON APIs and llms.txt. Agents can search servers and observed tools, retrieve grades and Trust Receipts, and search available evidence.