Local-first RAG MCP server: hybrid search over a folder of your own documents
Minirag-mcp is an MCP integration. Local-first RAG MCP server: hybrid search over a folder of your own documents.
The published pypi package is minirag-mcp, version 0.6.1. Its declared MCP transport is stdio.
The package declares BASE_DIR: One document root; also the security boundary for file access. Defaults to the process working directory. Ignored when BASE_DIRS is set.; BASE_DIRS: JSON array of document roots, e.g. ["/docs/a", "/docs/b"]. Takes precedence over BASE_DIR. An invalid value is a hard configuration error.; DB_PATH: LanceDB index directory. Defaults to /.minirag/lancedb, so each corpus gets its own index.; CACHE_DIR: Embedding model cache. Defaults to the platform user cache dir, so the ~220 MB model is downloaded once and shared.; MODEL_NAME: fastembed model id. Changing it makes existing vectors incompatible with new queries; pair with a new DB_PATH or a full re-ingest.; MAX_FILE_SIZE: Per-file size limit in bytes, enforced before parsing.; CHUNK_TOKEN_BUDGET: Retrieval-unit size in the embedding model's own tokens. Range 16-128; the ceiling is the model's trained sequence length.; RAG_HYBRID_WEIGHT: Keyword weight in the weighted RRF fusion, range 0.0-1.0. 0 is vector-only; higher values raise the BM25 contribution.; RAG_GROUPING: Result grouping filter. 'similar' keeps only the closest group; 'related' also keeps the next one. Unset means no grouping filter.; RAG_MAX_DISTANCE: Drop results whose vector distance exceeds this value. Lower is stricter. Unset means no distance filter.; RAG_MAX_FILES: Keep chunks from at most this many best-scoring files. Unset means no per-file filter.; RAG_INSTRUCTIONS_APPEND: Extra paragraph appended to the instructions the server hands the client at connect time, for corpus-specific guidance.; ALLOW_PRIVATE_URLS: Let ingest_url fetch hosts resolving to loopback, link-local, private, reserved or unspecified addresses. Off by default..
Check the maintainer’s instructions for account access, authentication and the supported client configuration. Once connected, inspect the tools your client receives and try a small task that matches the server’s stated purpose. The registry description does not include a verified, complete tool list.
View the publisher’s registry entry · Project repository · Project website. This profile uses version 0.6.1 from the registry. Registration identifies the publisher namespace; it does not establish that this is a first-party integration for every service it connects.
Browse related developer tools MCP servers for tools that cover the same kind of work.
| Variable | Description | Required |
|---|---|---|
| BASE_DIR | One document root; also the security boundary for file access. Defaults to the process working directory. Ignored when BASE_DIRS is set. | Optional |
| BASE_DIRS | JSON array of document roots, e.g. ["/docs/a", "/docs/b"]. Takes precedence over BASE_DIR. An invalid value is a hard configuration error. | Optional |
| DB_PATH | LanceDB index directory. Defaults to /.minirag/lancedb, so each corpus gets its own index. | Optional |
| CACHE_DIR | Embedding model cache. Defaults to the platform user cache dir, so the ~220 MB model is downloaded once and shared. | Optional |
| MODEL_NAME | fastembed model id. Changing it makes existing vectors incompatible with new queries; pair with a new DB_PATH or a full re-ingest. | Optional |
| MAX_FILE_SIZE | Per-file size limit in bytes, enforced before parsing. | Optional |
| CHUNK_TOKEN_BUDGET | Retrieval-unit size in the embedding model's own tokens. Range 16-128; the ceiling is the model's trained sequence length. | Optional |
| RAG_HYBRID_WEIGHT | Keyword weight in the weighted RRF fusion, range 0.0-1.0. 0 is vector-only; higher values raise the BM25 contribution. | Optional |
| RAG_GROUPING | Result grouping filter. 'similar' keeps only the closest group; 'related' also keeps the next one. Unset means no grouping filter. | Optional |
| RAG_MAX_DISTANCE | Drop results whose vector distance exceeds this value. Lower is stricter. Unset means no distance filter. | Optional |
| RAG_MAX_FILES | Keep chunks from at most this many best-scoring files. Unset means no per-file filter. | Optional |
| RAG_INSTRUCTIONS_APPEND | Extra paragraph appended to the instructions the server hands the client at connect time, for corpus-specific guidance. | Optional |
| ALLOW_PRIVATE_URLS | Let ingest_url fetch hosts resolving to loopback, link-local, private, reserved or unspecified addresses. Off by default. | Optional |
Official Redis server — search keys, inspect data structures and manage caches in natural language.
Enables AI agents and LLMs to interact with the web and extract information from web pages via the RAG Web Browser Actor.
Enables semantic search and retrieval of documentation using a vector database, allowing users to add and query documentation from URLs or local
Augments AI responses with relevant documentation context through vector search.
Augment AI responses with context by retrieving and processing documentation through vector search.
Enables semantic analysis of chat conversations through vector embeddings and knowledge graphs.