GalaxyAI setup

guide.md

Run FastQC tool

1. Install FastQC tool via the Admin panel (or on usegalaxy.org)
2. Register and name your History (e.g., FastQC workshop)
3. Upload a FASTQ dataset: https://zenodo.org/records/17929063/files/normal_R1.fastq.gz
4. Search for FastQC and run it
5. Open the HTML report (if on your VM, you'll need to add fastQC to sanitize allow list)

Configure GalaxyAI to use the Jetstream2 LLM service

0. Get your Jetstream2 LLM token from https://llm.jetstream-cloud.org/ (Settings -> Account -> API keys)
1. On the Galaxy VM, stop Galaxy if it is currently running.
2. Open the Galaxy configuration file:

vi config/galaxy.yml

3. Under the existing galaxy: section, add the Jetstream2 LLM settings. Replace <JS2-LLM-TOKEN> with the API token created in Open WebUI.

galaxy:
  admin_users: your_email@address
  inference_services:
    default:
      api_key: <YOUR-JS2-LLM-TOKEN>
      api_base_url: https://llm.jetstream-cloud.org/api/
      model: gpt-oss-120b
      temperature: 0.2

4. Save the file and restart Galaxy:

sh run.sh

5. After Galaxy starts, open the GalaxyAI chat surface and send a small test prompt:

Summarize what is currently in this Galaxy history.

Connect an External Client to Galaxy MCP

Before configuring your client to connect to Galaxy's MCP, type the following prompt:

Give me a 1-2 sentence description of Galaxy.

Claude Desktop configuration

{
  "mcpServers": {
    "galaxy-js2": {
      "command": "uvx",
      "args": ["galaxy-mcp"],
      "env": {
        "GALAXY_URL": "http://instance_ip/",
        "GALAXY_API_KEY": "<your-galaxy-api-key>"
      }
    }
  }
}

Codex app configuration

Add the following to the bottom of ~/.codex/config.toml

[mcp_servers.galaxy-js2]
command = "uvx"
args = ["galaxy-mcp"]

[mcp_servers.galaxy-js2.env]
GALAXY_URL = "http://<instance_ip>/"
GALAXY_API_KEY = "<your-galaxy-api-key>"

Try that initial prompt now:

Give me a 1-2 sentence description of Galaxy.

Some more sample prompts to try:

Can you connect to Galaxy and summarize my recent histories?

Find a tool named FastQC and explain what inputs it expects before running anything.

Using my current Galaxy history, identify the FASTQ dataset and prepare to run FastQC. Before executing, tell me exactly what tool and input you selected.