1
Create an API key
2
Install the CLI
tracks health prints ok when the API is reachable. It doesn’t check your key or your balance: the next step does.3
Start a run
id: the sandbox is still preparing.Without --budget or --time-limit, a run uses the default budget and time limit.4
Follow it
5
Read the result
6
See it in the web app
Open Runs in the web app and click your run. You see the agent’s conversation, the tools it called and the run’s settings, with its cost.
The sandbox is destroyed when the run ends. Anything the agent should keep, it must push or send somewhere, for
example by asking it in the prompt to open a pull request.
Compare two agents
One run tells you what one agent did. To know which agent does your task best, give the same task to several agents in an experiment: see compare two agents.Next steps
- Runs: follow-up prompts, setup scripts, files and limits.
- Private repositories: give the agent code that isn’t public.
- CLI reference and API reference.