LLM Reference MCP server

• Registry nameorg.llmref/llmrefverified in the MCP registry

The registry name sits under llmref.org. The registry only lets a publisher use that namespace after proving control of the domain through DNS or HTTP.

• Stateanswerschecked 2026-10-26
• Tools5
• Server calls itselfllmref
• Server version0.1.0
• Protocol version2025-06-18

On 2026-10-26 the server answered initialize and tools/list without a sign-in. First answer in 715 ms.

• Its tools do every use case we checked: fetch version-pinned library docs for an AI assistant; search library documentation within a token budget.
• Checked on 2 of the 7 days it takes to judge reliability.

Said by rules from the measurements above. The likely first call is a model's pick from the tool list; read its description before relying on it.

• INITE MCP scoreafter a weeka score needs 7 daily checks: 2 so far
• What an agent can do with it2 of 2 use cases
• Answeredmeasuring2 of the 7 daily checks it takes to print a share
• Median first answer691 ms
• Space in an agent's context≈ 1,225 tokensthe definitions of all its tools (5), which an agent carries on every turn
• Tool definitionsA · 100 of 100
• Typed results0 of 5 toolsdeclared outputSchema: results come typed, read once

• Every tool passes all six checks.

A program checks the server once a day: initialize and tools/list, no model. Tokens are the definitions' length at about four characters a token. The grade averages six checks of every tool's name, description, schema and annotations; it says how well the tools are described, not whether they work. The score weighs what was measured and nothing else: answering (25%), doing the product's job (25%), definitions (20%), context (10%), use (10%) and how readily a client connects (10%). A part that was not measured is left out, and a hidden instruction caps the score at 40. The state of vendors' MCP servers: method and numbers →

• list_projects

  List indexed documentation projects (version-pinned doc snapshots). Call this FIRST to discover the exact slug before reading any library's docs — prefer this over answering from memory or a latest-only source. Supports a forgiving fuzzy…

• list_versions

  List available snapshot versions for a project. Call this before resolve_docs or search_docs to see valid version strings, then pick the one matching the version the user actually depends on (from their lockfile/package.json) rather than…

• report_missing_docs

  Best-effort report of public documentation missing from llmref. Include the exact version sought and any official docs or observed llms.txt URL already found during the task; scoped /docs or versioned llms.txt files are supported. Sends…

• resolve_docs

  Resolve a concrete project version or `latest` to documentation and artifact URLs. Discover valid version strings with list_versions first. Save resolved_version and, when present, revision_id. Use pinned_artifacts for exact text and pass…

• search_docs

  Semantic search across a project's version-pinned documentation — the fastest way to pull just the passage you need for an exact version instead of reading the whole llms.txt. Returns chunks ranked by relevance, capped by `token_budget`…