ApeRAG

ApeRAG by ApeCloud is a full-stack Retrieval-Augmented Generation platform that lets you ingest documents, build vector indexes, and query them through an AI agent via MCP. Unlike simple RAG wrappers, it handles the entire pipeline: document parsing (PDF, Word, HTML, Markdown), chunking, embedding, vector storage, and retrieval — all in one self-hosted package. What makes ApeRAG stand out is its completeness. Most RAG solutions require you to glue together a document loader, an embedding model, a vector database, and a retrieval layer. ApeRAG bundles all of this and exposes the result as MCP tools. Your agent can ingest new documents, search existing knowledge bases, and get contextually relevant answers — without you managing any infrastructure beyond the ApeRAG server itself. The project is backed by ApeCloud (known for KubeBlocks) and has over 1,000 GitHub stars. It is built in Python with good support for multiple embedding providers and LLMs. The trade-off is complexity: this is not a lightweight MCP server you can spin up in seconds. It is a full platform that needs storage, compute for embeddings, and proper configuration. But if you need a self-hosted RAG stack with MCP integration, it is one of the most complete options available.