We help AI engineers and ML researchers evaluate their models and build environments!

This quickstart will allow you to:

1. Pull down an evaluation set
2. Initialize an eval environment on our hosted infra
3. Run an agent through a task on the environment
4. Evaluate the environment
5. Close the remote environment
6. View the trajectory and telemetry in our trace viewer (steps, tool-calls, score, and logs)

• pyproject.toml - dependencies for the hud-python SDK
• quickstart.py - a simple walkthrough of the full agent loop
• .env.example - a template for your environment variables
• README.md - this file

1. Sign up at app.hud.so
2. Get your API key from app.hud.so/project/api-keys
3. Configure environment variables - Copy .env.example to .env and add your:
   • HUD API key
   • OpenAI API key
   • Claude API key
4. Run the quickstart: cd quickstart && uv run quickstart.py