io.github.sfrangulov/minirag-mcp

{
  "_meta": {
    "io.modelcontextprotocol.registry/official": {
      "status": "active",
      "isLatest": true,
      "updatedAt": "2026-10-22T17:14:05.87471Z",
      "publishedAt": "2026-10-22T17:14:05.87471Z",
      "statusChangedAt": "2026-10-22T17:14:05.87471Z"
    }
  },
  "server": {
    "name": "io.github.sfrangulov/minirag-mcp",
    "$schema": "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json",
    "version": "1.0.0",
    "packages": [
      {
        "version": "1.0.0",
        "transport": {
          "type": "stdio"
        },
        "identifier": "minirag-mcp",
        "runtimeHint": "uvx",
        "registryType": "pypi",
        "registryBaseUrl": "https://pypi.org",
        "environmentVariables": [
          {
            "name": "BASE_DIR",
            "format": "filepath",
            "description": "One document root; also the security boundary for file access. Defaults to the process working directory. Ignored when BASE_DIRS is set."
          },
          {
            "name": "BASE_DIRS",
            "format": "string",
            "description": "JSON array of document roots, e.g. [\"/docs/a\", \"/docs/b\"]. Takes precedence over BASE_DIR. An invalid value is a hard configuration error.",
            "placeholder": "[\"/docs/a\", \"/docs/b\"]"
          },
          {
            "name": "DB_PATH",
            "format": "filepath",
            "description": "LanceDB index directory. Defaults to <first root>/.minirag/lancedb, so each corpus gets its own index."
          },
          {
            "name": "CACHE_DIR",
            "format": "filepath",
            "description": "Embedding model cache. Defaults to the platform user cache dir, so the ~220 MB model is downloaded once and shared."
          },
          {
            "name": "MODEL_NAME",
            "format": "string",
            "default": "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2",
            "description": "fastembed model id. Changing it makes existing vectors incompatible with new queries; pair with a new DB_PATH or a full re-ingest."
          },
          {
            "name": "MAX_FILE_SIZE",
            "format": "number",
            "default": "104857600",
            "description": "Per-file size limit in bytes, enforced before parsing."
          },
          {
            "name": "CHUNK_TOKEN_BUDGET",
            "format": "number",
            "default": "110",
            "description": "Retrieval-unit size in the embedding model's own tokens. Range 16-128; the ceiling is the model's trained sequence length."
          },
          {
            "name": "RAG_HYBRID_WEIGHT",
            "format": "number",
            "default": "0.6",
            "description": "Keyword weight in the weighted RRF fusion, range 0.0-1.0. 0 is vector-only; higher values raise the BM25 contribution."
          },
          {
            "name": "RAG_GROUPING",
            "format": "string",
            "choices": [
              "similar",
              "related"
            ],
            "description": "Result grouping filter. 'similar' keeps only the closest group; 'related' also keeps the next one. Unset means no grouping filter."
          },
          {
            "name": "RAG_MAX_DISTANCE",
            "format": "number",
            "description": "Drop results whose vector distance exceeds this value. Lower is stricter. Unset means no distance filter."
          },
          {
            "name": "RAG_MAX_FILES",
            "format": "number",
            "description": "Keep chunks from at most this many best-scoring files. Unset means no per-file filter."
          },
          {
            "name": "RAG_INSTRUCTIONS_APPEND",
            "format": "string",
            "description": "Extra paragraph appended to the instructions the server hands the client at connect time, for corpus-specific guidance."
          },
          {
            "name": "ALLOW_PRIVATE_URLS",
            "format": "boolean",
            "default": "false",
            "description": "Let ingest_url fetch hosts resolving to loopback, link-local, private, reserved or unspecified addresses. Off by default."
          }
        ]
      }
    ],
    "repository": {
      "id": "1327083343",
      "url": "https://github.com/sfrangulov/minirag-mcp",
      "source": "github"
    },
    "websiteUrl": "https://github.com/sfrangulov/minirag-mcp#readme",
    "description": "Local-first RAG MCP server: hybrid search over a folder of your own documents"
  }
}