Docs
Everything you need to train your first agent.
Start with one job and real examples.
Start
- Getting started Prepare one job and real examples, ask about early access, then run a small training session in Spark Loops.
- Your first agent What is inside one agent, what examples to gather, and why every new one starts in draft-only mode.
- How teams use Spark Choose a repeated job, gather examples and keep a person responsible for approving changes.
Train
- Running a training run Set a spending limit, choose what to work on and follow a training run in Spark Loops.
- Reading results Review a training run, read the proposed instructions and decide whether to approve a change.
- When progress stalls Review your examples and try further instruction variations when a training run stops improving.
- Real outcomes Use real outcomes to understand how your agent performs beyond a test.