Docs
A coding agent can run your evaluations for you. Copy the prompt below into your agent, then follow along.
Paste this prompt into Claude Code, Codex or another coding agent. The agent asks you to sign in, and you approve each purchase.
Help me evaluate my robot policy on Positronic Platform.
1. Run: uv tool install --managed-python --python 3.13 positronic. Then read the guide of the version that uv tool list shows: https://github.com/Positronic-Robotics/positronic/blob/v<version>/client/README.md
2. Run positronic account register in the background. Show me its sign-in line at once and wait while I sign in. Never print my API key. If it says that I have a key already, use the key saved on this computer, or ask me to set POSITRONIC_PLATFORM_API_KEY and restart you. Rotate the key only on my go.
3. Check my balance. Buy only the credits that I choose: show me checkout_url and wait while I pay. Never ask me for card details. If a purchase shows review_reason, stop and tell me. I can also ask [email protected] for welcome credits.
4. Push my policy image to a registry. Submit it with positronic eval run --org=<your org>, and check the status until the evaluation ends.- 1. Sign in
The agent installs the
positroniccommand line tool and starts the sign-in. You sign in in your browser. The tool saves your API key on your computer. - 2. Add credits
The agent checks your balance. You choose each purchase and pay for it yourself.
- 3. Submit
The agent pushes your policy image to a registry, submits the evaluation with
positronic eval run, and checks its status until it ends.
The guide to the positronic command line tool and its Python client is on GitHub: client/README.md. It covers the sign-in, credits, eval plans, the commands, their configuration and the Python client.