{"id":"ref-tools-mcp","name":"Ref Tools MCP","homepage":"https://ref.tools","repo_url":"https://github.com/ref-tools/ref-tools-mcp","category":"developer-tools","subcategories":["mcp-server","documentation","context-optimization"],"tags":["documentation","mcp","context-optimization","api-docs","coding-agent","token-efficiency","private-libraries"],"what_it_does":"MCP server that gives coding agents token-efficient access to public and private API/library documentation, filtering results to the most relevant ~5k tokens per query and deduplicating searches within sessions to prevent context rot.","use_cases":["Give coding agents access to third-party API documentation without bloating context with irrelevant pages","Search private library or internal documentation not indexed by default in LLM training data","Reduce LLM API costs by eliminating unnecessary documentation tokens during agentic coding loops","Provide a web URL reader that converts pages to clean markdown for agent consumption","Keep agent sessions on track by filtering previously seen search results within a session"],"not_for":["General web browsing or non-documentation content retrieval","Offline or air-gapped environments (requires API key and internet access)","Teams who need self-hosted documentation indexing without a SaaS dependency"],"best_when":"You have coding agents that frequently look up API and library documentation and are experiencing context bloat or degraded model performance from irrelevant documentation chunks.","avoid_when":"Your agents work with a fixed, small set of well-known libraries already in the model's training data, or you need fully offline documentation access.","alternatives":["Context7 MCP","DevDocs","direct WebFetch in agent prompts","llms.txt integrations"],"af_score":74.1,"security_score":68.0,"reliability_score":null,"package_type":"mcp_server","discovery_source":["github"],"priority":"low","status":"evaluated","version_evaluated":"latest","last_evaluated":"2026-03-01T09:50:06.132873+00:00","performance":{"latency_p50_ms":null,"latency_p99_ms":null,"uptime_sla_percent":null,"rate_limits":null,"data_source":"llm_estimated","measured_on":null}}