MCP Server Dagster

🚀 mcp-dagster: A Dagster MCP Server

The Model Context Protocol (MCP) is an open protocol that enables LLM applications to seamlessly integrate with external data sources and tools. This repository provides an MCP server for interacting with Dagster, a data orchestration platform.

🚀 Quick Start

Run Examples

1. Start a Dagster instance with your pipeline:

uv run dagster dev -f ./examples/open-ai-agent/pipeline.py

2. Run the MCP server using SSE (Server-Sent Events) transport:

uv run examples/open-ai-agent/run_sse_mcp.py

3. Start the agent loop that interacts with Dagster:

uv run ./examples/open-ai-agent/agent.py

Example Interactions

Once the agent is running, you can ask questions like:

• "What are the available assets in my Dagster instance and what do they do?"
• "Can you materialize the continent_stats asset and show the results?"
• "Check the status of recent runs and provide a summary of any failures."
• "Create a new monthly aggregation asset that depends on continent_stats."

The agent will use the MCP server to interact with your Dagster instance and provide answers based on your data pipeline.

✨ Features

This is a server that implements the Model Context Protocol, enabling AI agents to interact with Dagster instances to explore data pipelines, monitor runs, and manage assets. It serves as a bridge between LLMs and your data engineering workflows.

Read our launch article for more information.

📦 Installation

Not provided in the original document, so this section is skipped.

💻 Usage Examples

Basic Usage

The server implements several tools for interacting with Dagster:

• list_repositories: List all available Dagster repositories.
• list_jobs: List all jobs in a specific repository.
• list_assets: List all assets in a specific repository.
• recent_runs: Get recent Dagster runs (default limit is 10).
• get_run_info: Get detailed information about a specific run.
• launch_run: Launch a Dagster job run.
• materialize_asset: Materialize a specified Dagster asset.
• terminate_run: Terminate an ongoing Dagster run.
• get_asset_info: Get detailed information about a specific asset.

📚 Documentation

The server uses the following default configuration to connect to Dagster:

• GraphQL endpoint: http://localhost:3000/graphql
• Transport: SSE (Server-Sent Events)

🔧 Technical Details

Not provided in the original document, so this section is skipped.

📄 License

Not provided in the original document, so this section is skipped.