Each challenge adds a new power to a single agent you'll own by the end. Your agent must interact with our live API to pass — no static code review.
Fill in the starter code — tool definitions, system prompt, and the agentic loop.
Your agent calls our live API endpoints. Watch every tool call appear in the mission log in real time.
The key is hidden in an API response. When your agent calls submit_key() with the correct value, the challenge passes and the next one unlocks.
Want to start over? Resetting clears all progress from this browser.