gnosis-mcp
Gnosis MCP is a zero-config search tool for local documentation. It indexes markdown, git commits, web pages, and more into an SQLite database with FTS5 indexing, enabling fast keyword and semantic search without external APIs or cloud dependencies.
Ficha actualizada 23 sept 2026, 16:40 · upstream 23 sept 2026, 16:12 · release v0.17.5
Qué es gnosis-mcp
Descripción editorial construida a partir del repositorio, su documentación y sus señales de mantenimiento.
Gnosis MCP eliminates the need to paste entire files into LLM context by providing a searchable index of your local documentation. Instead of dumping thousands of tokens from full documents, it returns concise, relevant excerpts typically between 300–800 tokens. This reduces token costs by 5–10× while maintaining high recall through hybrid keyword and semantic search powered by ONNX embeddings. The tool supports multiple document formats including Markdown, Git commit history, web crawls via sitemap or link crawling, and various data types like CSVs, JSONs, and TOML files.
- Search your documentation for specific topics without pasting entire files into the LLM context
- Retrieve relevant code snippets from git commit history instead of searching raw source files
- Find configuration options or API endpoints by querying a crawled website's documentation
- Answer developer questions about project structure, conventions, and best practices
- Search through CSV data exports for specific values without loading the full dataset
- Query generated reports or notes to find relevant context quickly
- Zero-config deployment with SQLite by default; no external dependencies required
- Hybrid search combining BM25 keyword matching with ONNX embeddings for semantic relevance
- Measured performance: 8.7ms mean MCP round-trip time and 0.92 Hit@5 on real dev docs
- Built-in evaluation harness that produces reproducible metrics against a fixed fixture set
- Git history ingestion allows searching commit messages as searchable context
- Web crawling via sitemap or link extraction for indexing external documentation sources
- SQLite FTS5 is not available on all platforms (e.g., macOS) without additional compilation steps
- Embedding quality depends on the ONNX model used; results are not guaranteed to be perfect
- No built-in version control or change detection — you must manually track document updates
- Web crawling requires manual configuration of URLs and sitemap extraction logic
- Large-scale deployments (thousands of documents) may benefit from PostgreSQL with pgvector for vector search
Capturas del proyecto
Imágenes reales detectadas en la documentación del repositorio.
Perfil de infraestructura
Estimación inicial derivada del repositorio y de los servicios detectados.
Despliegue detectado
Qué señales reales de instalación hemos encontrado.
Install via pip install gnosis-mcp or uv tool install. Serve at http://localhost:8000 using docker-compose: docker compose up -d. For production, use systemd service files and configure environment variables like GNOSIS_MCP_ACCESS_LOG for audit trails. Enable FTS5 on SQLite by setting GNOSIS_MCP_FTS5=1 or compiling the project with FTS support. Always validate with gnosis-mcp check before connecting to your LLM client.
Stack y dependencias
Componentes inferidos a partir de manifests, configuración y documentación.
Puede sustituir a
Relaciones detectadas por el análisis del repositorio y su categoría.
Experiencia real de la comunidad
Ayuda a contrastar la estimación automática con instalaciones reales.
Todavía no hay datos comunitarios para este proyecto.