knowledge-rag
knowledge-rag is an MCP server. Its repository has 301 GitHub stars and is licensed under MIT; we last read it on 2026-10-22.
knowledge-rag provides a local RAG (Relevant Answer Generation) server for Claude Code, integrating hybrid search methods (semantic and BM25) with cross-encoder reranking, offering 13 MCP tools and 20 format parsers. No external servers or API keys are required.
Install
Read from knowledge-rag’s own README, close to verbatim -- not written by us, and not guessed.
pip install knowledge-rag → restart Claude Code → search_knowledge("your query")What it reaches, and what it needs
What it connects to
Unverified means a mention in the server’s own description, not a citation. A wrong claim here costs nothing to correct and nothing to trust yet.
Maintenance
Local RAG MCP server for Claude Code — hybrid search (semantic + BM25), cross-encoder reranking, 13 MCP tools, 20 format parsers. Zero external servers, zero API keys.
What knowledge-rag can do
10 tools documented in its README. An AI client calls these by name.
search_knowledge- No description in the README.
rag-cite-sources- Every claim ships with path:line citations.
rag-onboard-context- First interaction of a session probes what is indexed.
rag-deep-dive- 3-step drill: search → fetch → find similar.
rag-web-fallback- Only hit the web when local RAG comes back empty.
rag-troubleshoot- Bug / error → RAG first for prior fixes.
rag-code-review- Review consults ADRs / patterns before commenting.
rag-index-decisions- After a decision, index it back — close the feedback loop.
rag-security-first- Security tasks: MITRE / CVE / runbook first.
rag-evaluate-quality- Weekly checkup — MRR@5 · Recall@5 · Precision@5.
Read from the README; names are kept only if the README contains them.
Common questions about knowledge-rag
- What does knowledge-rag do?
- knowledge-rag provides a local RAG (Relevant Answer Generation) server for Claude Code, integrating hybrid search methods (semantic and BM25) with cross-encoder reranking, offering 13 MCP tools and 20 format parsers. No external servers or API keys are required.
- How do I install knowledge-rag?
- Its own README gives this command: pip install knowledge-rag → restart Claude Code → search_knowledge("your query"). Run it where your MCP client can launch the server, and check the README for any settings it asks for.
- Which AI clients can use knowledge-rag?
- Its own description mentions Claude Code. That is a mention in its description, not a test we ran.
- Is knowledge-rag official and is it maintained?
- We found nothing naming it as a vendor’s official server. Its repository has 301 GitHub stars and its state in our directory is listed. Its last commit we recorded was on 2026-10-16.
- What licence does knowledge-rag use?
- The repository declares the MIT licence.