Skip to content

Get Started

Choose the path that matches how you want to run models. Both begin with the same package and take you to a verified first result.

CLI track

Use this track when you want to orchestrate coding-agent CLIs such as Codex or Claude Code from the terminal.

  1. Install LionAGI and authenticate a CLI provider.
  2. Run li doctor.
  3. Run one agent, continue it, then preview orchestration.

Your first success is a non-empty agent response plus the [to resume] hint printed by li agent.

Python track

Use this track when LionAGI is a dependency in your application and you want API-backed models, typed output, or programmatic tool orchestration.

  1. Install LionAGI in a project.
  2. Export one provider API key.
  3. Record a chat turn and request typed output.

Your first success is a recorded Branch.communicate() response followed by a validated Pydantic model from Branch.operate().

Not sure which path you need?

Use the surface chooser. It selects the lightest surface from the shape of the task, without requiring you to learn the full architecture first.