deepset-mcp

The official MCP server and Python SDK for the Haystack Enterprise Platform

deepset-mcp enables AI agents to build and debug pipelines on the Haystack Enterprise Platform through 45+ specialized tools. It also provides a Python SDK for programmatic access to many platform resources.

Documentation

📖 View the full documentation

Quick Links

• 🔗 Haystack Enterprise Platform
• 📚 Installation Guide
• 🛠️ MCP Server Guide
• 🐍 Python SDK Guide

Development

Installation

Install the project using uv:

# Install uv first
pipx install uv

# Install project with all dependencies
uv sync --locked --all-extras --all-groups

Local Development

If you want to test your changes locally, follow these steps:

1. Add a script run-deepset-mcp.sh that uses the binary from the project's virtual env

#!/usr/bin/env bash
# Wrapper to run the local deepset-mcp server for Cursor MCP.
# Use this as command so it doesn't depend on uv or PATH.
set -e
cd "$(dirname "$0")"
exec .venv/bin/deepset-mcp

2. Use it this way in Cursor:

    "deepset": {
      "command": "/bin/bash",
      "args": ["/Users/*****/****/deepset-mcp-server/run-deepset-mcp.sh"],
      "cwd": "/Users/*****/****/deepset-mcp-server",
      "env": {
        "DEEPSET_WORKSPACE": "WORKSPACE",
        "DEEPSET_API_KEY": "API_KEY"
      }
    }

Note: If you change the codebase, make sure to restart the MCP server.

Code Quality & Testing

Run code quality checks and tests using the Makefile:

# Install dependencies
make install

# Code quality
make lint          # Run ruff linting
make format        # Format code with ruff
make types         # Run mypy type checking

# Testing
make test          # Run unit tests (default)
make test-unit     # Run unit tests only
make test-integration     # Run integration tests
make test-all      # Run all tests

# Clean up
make clean         # Remove cache files

Documentation

Documentation is built using MkDocs with the Material theme:

• Configuration: mkdocs.yml
• Content: docs/ directory
• Auto-generated API docs via mkdocstrings
• Deployed via GitHub Pages (automated via GitHub Actions on push to main branch)