Skip to content

deepset-mcp

The official MCP server and Python SDK for the deepset AI platform

deepset-mcp provides two powerful layers for interacting with the deepset AI platform:

  • MCP Server: Enables AI agents to build and debug pipelines on the deepset platform through 30+ specialized tools
  • Python SDK: Provides programmatic access to many deepset platform resources for developers

Quick Start

Get started with the MCP server in 2 minutes:

  1. Install uv (Python package manager): 

    pipx install uv

  2. Configure your MCP client (Cursor example): 

    {
  "mcpServers": {
    "deepset": {
      "command": "uvx",
      "args": ["deepset-mcp"],
      "env": {
        "DEEPSET_WORKSPACE": "your-workspace",
        "DEEPSET_API_KEY": "your-api-key"
      }
    }
  }
}

  3. Start building: "Create a RAG pipeline with hybrid retrieval using Claude Sonnet as the LLM"

What You Can Do

With the MCP Server:

  • Create and deploy AI pipelines through natural language

  • Debug pipeline issues with intelligent log analysis

  • Manage indexes, templates, and workspace resources

  • Test pipelines interactively through agents

With the Python SDK:

  • Programmatically manage pipelines, indexes, and workspaces

  • Build custom tooling on top of deepset platform APIs

  • Integrate deepset capabilities into existing applications

  • Automate deployment and management workflows

Documentation Structure

Installation

Set up deepset-mcp with Cursor, Claude Desktop, or other MCP clients.

Guides

  • MCP Server: Configure and customize the MCP server
  • API SDK: Use the Python SDK programmatically

Concepts

Reference

Benefits

Faster Development: Build pipelines through conversation instead of clicking through UIs

Intelligent Debugging: Agents analyze logs and suggest fixes automatically

Flexible Access: Choose between agent-driven workflows (MCP) or direct API control (SDK)

Production Ready: Built by deepset with enterprise-grade reliability