Enterprise Internal Knowledge Base: Production-Ready RAG + MCP
Reference RAG pipeline exposed as an MCP server, with eval gating and tracing
- Pricing
- Open source
- Platforms
- CLI, API, self-hosted
- Version
- 0.1.0
- Listing updated
- today
Similar MCP servers
About Enterprise Internal Knowledge Base: Production-Ready RAG + MCP
Enterprise Internal Knowledge Base: Production-Ready RAG + MCP is an open-source MCP server for knowledge and memory. It is available on CLI, API and self-hosted.
The README describes this as a demo of retrieval-augmented generation run as an MCP server, built on a public corpus of Veterans Affairs education manuals. Documents are crawled, split with a structure-aware chunker that keeps table spans intact, embedded locally with mxbai-embed-large into Postgres with pgvector, and then retrieved and reranked at query time.
The server is built with FastMCP and offers a query tool, document resources and a citation prompt. Claude Desktop is the example client. Generation uses Claude Sonnet, and Langfuse traces are shared publicly for selected runs.
The repository focuses on evaluation and observability. A 110-question golden set is scored with DeepEval, and a GitHub Actions workflow blocks pull requests that regress past set thresholds. It suits developers who want to study or adapt a RAG pipeline with measurable quality checks. The README says tenant access control and cost tuning are covered in a linked article rather than in the code.
What it does
- MCP query tool
- Exposes ranked document chunks with citations to MCP clients such as Claude Desktop.
- Structure-aware chunking
- Splits source HTML and PDFs while preserving table colspan and rowspan structure.
- Hybrid reranking pipeline
- Reranks pgvector results with mxbai-rerank, or with a lighter FlashRank model in fast mode.
- Eval-gated merge checks
- Runs a golden query set in CI and blocks pull requests that exceed set regression thresholds.
- Langfuse tracing
- Sends per-query traces to Langfuse Cloud, with public sharing enabled for selected traces.
- Local embeddings
- Generates 1024-dimension embeddings locally with mxbai-embed-large, avoiding embedding API costs.
Good for
- Prototyping a retrieval-augmented MCP server over a document corpus
- Adding regression tests for RAG answer quality to a CI pipeline
- Tracing latency and retrieval behaviour across queries with Langfuse
- Studying chunking and reranking choices for government or manual-style documents
Pricing
Open source
Free MIT-licensed repository; running it requires Anthropic and Langfuse accounts.
Enterprise Internal Knowledge Base: Production-Ready RAG + MCP FAQ
What license is it released under?
The repository is released under the MIT license.
How do I connect it to Claude Desktop?
Add an entry under mcpServers in the Claude Desktop config that runs scripts/mcp_server.py with an absolute path to the repository.
Does the repository include multi-tenant access control?
No. The auth_context parameter is present on every MCP tool but is currently unused, and the README treats it as a seam for future SSO or ACL work.
What does the fixture database contain?
The README says it holds 238 Veterans Affairs education documents split into more than 9,000 chunks that are already embedded.
Is Enterprise Internal Knowledge Base: Production-Ready RAG + MCP free?
Yes. Enterprise Internal Knowledge Base: Production-Ready RAG + MCP is open source, so you can use and self-host it for free.
What are the best alternatives to Enterprise Internal Knowledge Base: Production-Ready RAG + MCP?
The closest alternatives to Enterprise Internal Knowledge Base: Production-Ready RAG + MCP are MiniRAG, rag-mcp, Chaos Cypher, cob-rag-mcp and MCP Server.
Where this comes from
- Description: written from the project's README and its MCP Registry entry.
