txtai exposes configured API endpoints as MCP tools for semantic search, retrieval pipelines and workflows through its Python API server.
txtai exposes configured API endpoints as MCP tools for semantic search, retrieval pipelines and workflows through its Python API server.
txtai is a Python framework for embeddings, retrieval, pipelines and workflows. Its API can export enabled endpoints as MCP tools. This is a real server configuration, rather than a fixed collection of tools supplied to every user: the API configuration and the resources loaded behind each endpoint determine the usable interface.
Add mcp: true to the API YAML configuration and start the configured API application. A dictionary can supply HTTP-client and FastApiMCP options when defaults are unsuitable. The resulting route is /mcp. Install the API dependencies and configure the index or pipelines needed by the task before connecting an assistant.
The documented implementation mounts the route through FastApiMCP's mount() interface, which uses SSE. Select a compatible client and inspect the installed dependency version before assuming that a generic HTTP connector will work. A local API URL is an example deployment address, not a public hosted txtai endpoint.
Build a small index with passages whose origin is known. Query it through the API and MCP, then compare the returned text and metadata with the source documents. Add a generation or workflow endpoint only when its input and result contract is clear. Configure API authentication and exposed endpoints deliberately; installing txtai alone does not activate every framework capability.
CONFIG=app.yml uvicorn "txtai.api:app"Create app.yml with mcp: true and the API resources needed for your task. This command starts the API; connect the client to your instance’s /mcp route.
Python 3.11 or later, txtai API dependencies, a configured index or pipeline and an SSE-compatible MCP client.
| Variable | Description | Required |
|---|---|---|
| CONFIG | Path to the API YAML configuration, including the MCP setting and enabled pipelines or index. | Yes |
Store entities, relations and observations in a local knowledge graph with the reference Memory MCP server.
A structured scratchpad for hard problems — stepwise reasoning with revisions, branches and visible logic.
This MCP server (Model Context Protocol) acts as a bridge between modern AI agents (Claude Desktop, Augment, Cursor) and the Thomson MO5 ecosystem. It allows access to an expert knowledge base and…
Downloads websites and their assets for use with Retrieval-Augmented Generation (RAG) systems.
Repository knowledge graph MCP server for codebase understanding and debugging.
Cross-tool persistent memory and context for AI assistants over MCP.