A Model Context Protocol (MCP) server designed for Nextflow development and testing, built with FastMCP.
Warning
This MCP is designed for developing Nextflow itself, not for writing Nextflow pipelines.
This MCP provides a suite of tools for Nextflow development, including:
- Building Nextflow from source
- Running tests (integration tests, specific tests, plugin tests)
- Running the development version of Nextflow
- Managing the Nextflow development directory
- Accessing Nextflow documentation
- Claude Desktop App
- Python 3.8+ with pip
- Git repository of Nextflow (cloned locally)
- Install the FastMCP CLI tool:
pip install fastmcp
- Clone this repository:
git clone https://github.com/yourusername/nextflow-dev-mcp.git
cd nextflow-dev-mcp
- Install the MCP in Claude Desktop:
fastmcp install mcp-nextflow.py
This will make the MCP available in the Claude Desktop app.
- Fetch the virtual environment path which includes the FastMCP CLI tool. If you are using
uv
this will be in the.venv
directory. - Get the directory of your Nextflow cloned repository.
- Add the following json to the cursor MCP servers:
{
"mcpServers": {
"server-name": {
"command": "/path/to/your/.venv/bin/python",
"args": [
"/path/to/your/mcp-nextflow/mcp-nextflow.py"
],
"env": {
"NEXTFLOW_DIR": "/path/to/your/nextflow"
}
}
}
}
Then, you should be able to use the MCP in Cursor. In Agentic mode, ask the agent to "test the nf-amazon plugin" and it should run make test module=plugins:nf-amazon
.
You can specify the Nextflow directory during installation:
NEXTFLOW_DIR=/path/to/your/nextflow fastmcp install mcp-nextflow.py
Once installed, you can access the MCP in the Claude Desktop app:
- Open Claude Desktop
- Click on the Tools menu button in the Claude interface
- Select Nextflow Developer Tools from the list of installed MCPs
Cursor is an AI-powered code editor that works with Claude. To use the MCP with Cursor:
- Make sure you've installed the MCP as described above
- Open your Nextflow project in Cursor
- In a chat with Claude in Cursor, you can reference the MCP:
Using the Nextflow Developer Tools, run the integration tests for the nf-amazon plugin