dagster-mcp 0.1.0

MCP server that wraps the Dagster GraphQL API — manage runs, assets, schedules, sensors, and backfills from any MCP client

Dagster MCP

An MCP server that wraps the Dagster GraphQL API, giving any MCP client (Claude Code, Cursor, etc.) full visibility and control over your Dagster instance.

What it does

Exposes up to 15 tools that let you inspect and operate a Dagster instance (self-hosted or Dagster Cloud):

• Runs — list, inspect, get logs, get step stats
• Assets — search, get details, get recent materializations
• Jobs — list all jobs across code locations
• Schedules & Sensors — list with status, cron, targets
• Code Locations — list and reload
• Backfills — list recent backfills
• Actions — launch jobs, terminate runs, reload code locations (opt-in, disabled by default)

Quick start

Prerequisites

• Python 3.12+
• uv (recommended) or pip
• A running Dagster instance (self-hosted or Cloud)

Install

Option A — run directly with uvx (no clone needed):

uvx dagster-mcp

Option B — clone and run:

git clone https://github.com/fabdendev/dagster-mcp.git
cd dagster-mcp
uv sync

Configure

Variable	Description	Default
DAGSTER_URL	Base URL of your Dagster instance	http://localhost:3000
DAGSTER_API_TOKEN	Dagster Cloud API token (leave empty for self-hosted)	(empty)
DAGSTER_READ_ONLY	When true, only read tools are exposed (no launch/terminate/reload)	true

Self-hosted:

export DAGSTER_URL=http://localhost:3000

Dagster Cloud:

export DAGSTER_URL=https://myorg.dagster.cloud/prod
export DAGSTER_API_TOKEN=your-dagster-cloud-user-token

Add to Claude Code

Add to your Claude Code MCP settings (~/.claude/settings.json):

If using uvx:

{
  "mcpServers": {
    "dagster": {
      "command": "uvx",
      "args": ["dagster-mcp"],
      "env": {
        "DAGSTER_URL": "http://localhost:3000",
        "DAGSTER_API_TOKEN": "",
        "DAGSTER_READ_ONLY": "true"
      }
    }
  }
}

If installed from clone:

{
  "mcpServers": {
    "dagster": {
      "command": "uv",
      "args": ["run", "--directory", "/path/to/dagster-mcp", "dagster-mcp"],
      "env": {
        "DAGSTER_URL": "http://localhost:3000",
        "DAGSTER_API_TOKEN": "",
        "DAGSTER_READ_ONLY": "true"
      }
    }
  }
}

Tools

Tool	Description
get_runs	List recent runs, filter by job name and/or status
get_run_status	Get status and info for a specific run
get_run_logs	Get logs/events for a run (with pagination)
get_run_stats	Get step-level stats (timing, materializations)
get_recent_materializations	Get recent materializations for an asset
get_asset_details	Get details, dependencies, and partitions for assets
search_assets	Search/list assets by key prefix or group name
list_jobs	List all jobs across all code locations
list_schedules	List schedules with status, cron, and next tick
list_sensors	List sensors with status and target jobs
list_code_locations	List all code locations and their load status
list_backfills	List recent backfills with status and progress
reload_code_location	Reload a code location (e.g. after deploy) (write)
terminate_run	Terminate a running run (write)
launch_job	Launch a job or materialize specific assets (write)

Tools marked (write) are only available when DAGSTER_READ_ONLY=false.

How it differs from the official Dagster MCP

The official Dagster MCP (dg[mcp]) is a development-time tool — it helps AI write Dagster code, scaffold components, and work with the dg CLI.

This project is an operations-time tool — it lets AI (or any MCP client) monitor and operate a running Dagster instance: inspect runs, read logs, check schedules, launch jobs, and more.

They serve different purposes and work well together.

Compatibility

Tested with Dagster 1.6+. All GraphQL queries target stable, non-deprecated API fields.

Development

uv sync --extra dev
uv run ruff check dagster_mcp/    # lint
uv run python -m dagster_mcp      # start server locally

License

MIT