--- bezeichner: "io.github.sfrangulov/minirag-mcp" art: "mcp_server" slug: "io-github-sfrangulov-minirag-mcp" paketkoordinate: "pypi:minirag-mcp" status: "aktiv" homepage: "https://github.com/sfrangulov/minirag-mcp#readme" erhebungsstand: "2026-10-10T01:17:01.464Z" namensraum: "io.github.sfrangulov" registerseite: "https://tracevero.com/mcp/io-github-sfrangulov-minirag-mcp" abgerufen_am: "2026-10-10" zugangsdaten_erforderlich: false ausfuehrungsort: "lokal" dateisystem_pfadargument: true quelloffen_einsehbar: true einsatzgebiet: "dokument" roh_beschreibung: "Local-first RAG MCP server: hybrid search over a folder of your own documents" version: "0.6.1" roh_umgebungsvariablen: "BASE_DIR, BASE_DIRS, DB_PATH, CACHE_DIR, MODEL_NAME, MAX_FILE_SIZE, CHUNK_TOKEN_BUDGET, RAG_HYBRID_WEIGHT, RAG_GROUPING, RAG_MAX_DISTANCE, RAG_MAX_FILES, RAG_INSTRUCTIONS_APPEND, ALLOW_PRIVATE_URLS" roh_geheime_pflichtvariablen: "" roh_transportarten: "stdio" roh_pfadargumente: "" roh_repository_url: "https://github.com/sfrangulov/minirag-mcp" roh_paketquellen: "pypi" roh_geheime_pflichtkopfzeilen: "" roh_pfad_umgebungsvariablen: "BASE_DIR, CACHE_DIR, DB_PATH" roh_remote_adressen: "" roh_remote_hosts: "" roh_statusmeldung: "" roh_veroeffentlicht_am: "2026-09-07" roh_aktualisiert_am: "2026-09-07" roh_schemafassung: "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json" roh_bereitstellungsform: "paket" roh_repository_quelle: "github" roh_repository_unterordner: "" roh_paketbezeichner: "minirag-mcp" roh_paketversionen: "0.6.1" roh_laufzeithinweise: "uvx" roh_umgebungsformate: "boolean, filepath, number, string" roh_umgebungsbeschreibungen: "ALLOW_PRIVATE_URLS=Let ingest_url fetch hosts resolving to loopback, link-local, private, reserved or unspecified addresses. Off by default. · 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. · CACHE_DIR=Embedding model cache. Defaults to the platform user cache dir, so the ~220 MB model is downloaded once and shared. · 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. · DB_PATH=LanceDB index directory. Defaults to /.minirag/lancedb, so each corpus gets its own index. · MAX_FILE_SIZE=Per-file size limit in bytes, enforced before parsing. · 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. · RAG_GROUPING=Result grouping filter. 'similar' keeps only the closest group; 'related' also keeps the next one. Unset means no grouping filter. · 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_INSTRUCTIONS_APPEND=Extra paragraph appended to the instructions the server hands the client at connect time, for corpus-specific guidance. · 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." roh_symbolformate: "" roh_verbindungswege: "{\"packages\":[{\"registryType\":\"pypi\",\"identifier\":\"minirag-mcp\",\"version\":\"0.6.1\",\"runtimeHint\":\"uvx\",\"transport\":\"stdio\",\"environment\":[{\"name\":\"BASE_DIR\",\"description\":\"One document root; also the security boundary for file access. Defaults to the process working directory. Ignored when BASE_DIRS is set.\",\"format\":\"filepath\",\"required\":false,\"secret\":false},{\"name\":\"BASE_DIRS\",\"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.\",\"format\":\"string\",\"required\":false,\"secret\":false},{\"name\":\"DB_PATH\",\"description\":\"LanceDB index directory. Defaults to /.minirag/lancedb, so each corpus gets its own index.\",\"format\":\"filepath\",\"required\":false,\"secret\":false},{\"name\":\"CACHE_DIR\",\"description\":\"Embedding model cache. Defaults to the platform user cache dir, so the ~220 MB model is downloaded once and shared.\",\"format\":\"filepath\",\"required\":false,\"secret\":false},{\"name\":\"MODEL_NAME\",\"description\":\"fastembed model id. Changing it makes existing vectors incompatible with new queries; pair with a new DB_PATH or a full re-ingest.\",\"format\":\"string\",\"required\":false,\"secret\":false},{\"name\":\"MAX_FILE_SIZE\",\"description\":\"Per-file size limit in bytes, enforced before parsing.\",\"format\":\"number\",\"required\":false,\"secret\":false},{\"name\":\"CHUNK_TOKEN_BUDGET\",\"description\":\"Retrieval-unit size in the embedding model's own tokens. Range 16-128; the ceiling is the model's trained sequence length.\",\"format\":\"number\",\"required\":false,\"secret\":false},{\"name\":\"RAG_HYBRID_WEIGHT\",\"description\":\"Keyword weight in the weighted RRF fusion, range 0.0-1.0. 0 is vector-only; higher values raise the BM25 contribution.\",\"format\":\"number\",\"required\":false,\"secret\":false},{\"name\":\"RAG_GROUPING\",\"description\":\"Result grouping filter. 'similar' keeps only the closest group; 'related' also keeps the next one. Unset means no grouping filter.\",\"format\":\"string\",\"required\":false,\"secret\":false},{\"name\":\"RAG_MAX_DISTANCE\",\"description\":\"Drop results whose vector distance exceeds this value. Lower is stricter. Unset means no distance filter.\",\"format\":\"number\",\"required\":false,\"secret\":false},{\"name\":\"RAG_MAX_FILES\",\"description\":\"Keep chunks from at most this many best-scoring files. Unset means no per-file filter.\",\"format\":\"number\",\"required\":false,\"secret\":false},{\"name\":\"RAG_INSTRUCTIONS_APPEND\",\"description\":\"Extra paragraph appended to the instructions the server hands the client at connect time, for corpus-specific guidance.\",\"format\":\"string\",\"required\":false,\"secret\":false},{\"name\":\"ALLOW_PRIVATE_URLS\",\"description\":\"Let ingest_url fetch hosts resolving to loopback, link-local, private, reserved or unspecified addresses. Off by default.\",\"format\":\"boolean\",\"required\":false,\"secret\":false}],\"additional_arguments_declared\":false}],\"remotes\":[]}" --- # io.github.sfrangulov/minirag-mcp ## Measured values | Property | Value | Source | Collected on | Level of trust | Raw declaration | | --- | --- | --- | --- | --- | --- | | Required secrets declared | false | MCP-Register | 2026-08-10T01:17:01.536Z | abgeleitet | roh_geheime_pflichtvariablen: ; roh_geheime_pflichtkopfzeilen: | | Execution location | lokal | MCP-Register | 2026-08-10T01:17:01.536Z | abgeleitet | roh_transportarten: stdio | | Path argument present | true | MCP-Register | 2026-08-16T01:17:01.420Z | abgeleitet | roh_pfadargumente: ; roh_pfad_umgebungsvariablen: BASE_DIR, CACHE_DIR, DB_PATH | | Repository URL listed | true | MCP-Register | 2026-08-10T01:17:01.536Z | abgeleitet | roh_repository_url: https://github.com/sfrangulov/minirag-mcp | | Field of use, derived from the vendor description | dokument | MCP-Register | 2026-08-26T16:45:01.514Z | abgeleitet | roh_beschreibung: Local-first RAG MCP server: hybrid search over a folder of your own documents | | Description (raw) | Local-first RAG MCP server: hybrid search over a folder of your own documents | MCP-Register | 2026-08-26T16:45:01.514Z | selbstauskunft | | | Declared version | 0.6.1 | MCP-Register | 2026-09-08T01:17:01.649Z | selbstauskunft | | | Environment variables (raw) | BASE_DIR, BASE_DIRS, DB_PATH, CACHE_DIR, MODEL_NAME, MAX_FILE_SIZE, CHUNK_TOKEN_BUDGET, RAG_HYBRID_WEIGHT, RAG_GROUPING, RAG_MAX_DISTANCE, RAG_MAX_FILES, RAG_INSTRUCTIONS_APPEND, ALLOW_PRIVATE_URLS | MCP-Register | 2026-08-10T01:17:01.536Z | selbstauskunft | | | Required secret variables (raw) | | MCP-Register | 2026-08-10T01:17:01.536Z | selbstauskunft | | | Transports (raw) | stdio | MCP-Register | 2026-08-10T01:17:01.536Z | selbstauskunft | | | Path arguments (raw) | | MCP-Register | 2026-08-10T01:17:01.536Z | selbstauskunft | | | Repository (raw) | https://github.com/sfrangulov/minirag-mcp | MCP-Register | 2026-08-10T01:17:01.536Z | selbstauskunft | | | Package registries (raw) | pypi | MCP-Register | 2026-08-10T01:17:01.536Z | selbstauskunft | | | Required secret headers (raw) | | MCP-Register | 2026-08-16T01:17:01.420Z | selbstauskunft | | | Path environment variables (raw) | BASE_DIR, CACHE_DIR, DB_PATH | MCP-Register | 2026-08-16T01:17:01.420Z | selbstauskunft | | | Remote URLs (raw) | | MCP-Register | 2026-08-16T01:17:01.420Z | selbstauskunft | | | Remote hosts (raw) | | MCP-Register | 2026-08-16T01:17:01.420Z | selbstauskunft | | | Registry status message (raw) | | MCP-Register | 2026-08-16T01:17:01.420Z | selbstauskunft | | | First listed in the registry (raw) | 2026-09-07 | MCP-Register | 2026-09-08T01:17:01.649Z | selbstauskunft | | | Last changed in the registry (raw) | 2026-09-07 | MCP-Register | 2026-09-08T01:17:01.649Z | selbstauskunft | | | Schema version of the raw record (raw) | https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json | MCP-Register | 2026-08-16T01:17:01.420Z | selbstauskunft | | | Delivery form (raw) | paket | MCP-Register | 2026-08-16T01:17:01.420Z | selbstauskunft | | | Repository platform (raw) | github | MCP-Register | 2026-08-16T01:17:01.420Z | selbstauskunft | | | Repository subfolder (raw) | | MCP-Register | 2026-08-16T01:17:01.420Z | selbstauskunft | | | Package identifiers (raw) | minirag-mcp | MCP-Register | 2026-08-16T01:17:01.420Z | selbstauskunft | | | Package versions (raw) | 0.6.1 | MCP-Register | 2026-09-08T01:17:01.649Z | selbstauskunft | | | Runtime hints (raw) | uvx | MCP-Register | 2026-08-16T01:17:01.420Z | selbstauskunft | | | Environment variable formats (raw) | boolean, filepath, number, string | MCP-Register | 2026-08-16T01:17:01.420Z | selbstauskunft | | | Environment variable descriptions (raw) | ALLOW_PRIVATE_URLS=Let ingest_url fetch hosts resolving to loopback, link-local, private, reserved or unspecified addresses. Off by default. · 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. · CACHE_DIR=Embedding model cache. Defaults to the platform user cache dir, so the ~220 MB model is downloaded once and shared. · 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. · DB_PATH=LanceDB index directory. Defaults to /.minirag/lancedb, so each corpus gets its own index. · MAX_FILE_SIZE=Per-file size limit in bytes, enforced before parsing. · 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. · RAG_GROUPING=Result grouping filter. 'similar' keeps only the closest group; 'related' also keeps the next one. Unset means no grouping filter. · 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_INSTRUCTIONS_APPEND=Extra paragraph appended to the instructions the server hands the client at connect time, for corpus-specific guidance. · 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. | MCP-Register | 2026-08-16T01:17:01.420Z | selbstauskunft | | | Icon formats (raw) | | MCP-Register | 2026-08-16T01:17:01.420Z | selbstauskunft | | | Connection paths (source structure) | {"packages":[{"registryType":"pypi","identifier":"minirag-mcp","version":"0.6.1","runtimeHint":"uvx","transport":"stdio","environment":[{"name":"BASE_DIR","description":"One document root; also the security boundary for file access. Defaults to the process working directory. Ignored when BASE_DIRS is set.","format":"filepath","required":false,"secret":false},{"name":"BASE_DIRS","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.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"LanceDB index directory. Defaults to /.minirag/lancedb, so each corpus gets its own index.","format":"filepath","required":false,"secret":false},{"name":"CACHE_DIR","description":"Embedding model cache. Defaults to the platform user cache dir, so the ~220 MB model is downloaded once and shared.","format":"filepath","required":false,"secret":false},{"name":"MODEL_NAME","description":"fastembed model id. Changing it makes existing vectors incompatible with new queries; pair with a new DB_PATH or a full re-ingest.","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Per-file size limit in bytes, enforced before parsing.","format":"number","required":false,"secret":false},{"name":"CHUNK_TOKEN_BUDGET","description":"Retrieval-unit size in the embedding model's own tokens. Range 16-128; the ceiling is the model's trained sequence length.","format":"number","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword weight in the weighted RRF fusion, range 0.0-1.0. 0 is vector-only; higher values raise the BM25 contribution.","format":"number","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Result grouping filter. 'similar' keeps only the closest group; 'related' also keeps the next one. Unset means no grouping filter.","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Drop results whose vector distance exceeds this value. Lower is stricter. Unset means no distance filter.","format":"number","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Keep chunks from at most this many best-scoring files. Unset means no per-file filter.","format":"number","required":false,"secret":false},{"name":"RAG_INSTRUCTIONS_APPEND","description":"Extra paragraph appended to the instructions the server hands the client at connect time, for corpus-specific guidance.","format":"string","required":false,"secret":false},{"name":"ALLOW_PRIVATE_URLS","description":"Let ingest_url fetch hosts resolving to loopback, link-local, private, reserved or unspecified addresses. Off by default.","format":"boolean","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]} | MCP-Register | 2026-09-08T01:17:01.649Z | selbstauskunft | | ## Links - Namespace: [[namensraum/io-github-sfrangulov|io.github.sfrangulov]] --- - Registry page: - Retrieved on: 2026-10-10