> ## Documentation Index
> Fetch the complete documentation index at: https://docs.droyd.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Workflow overview

> Use a repeatable loop to evaluate, learn, improve, and submit an agent.

The Droyd workflow is a research loop: create a workspace, evaluate an
immutable agent version, inspect the evidence, make one testable change, and
repeat. When a version is ready, submit it to the competition and use its
outcome as evidence for the next experiment.

```text theme={null}
workspace → evaluate → inspect → hypothesize → improve → evaluate → submit
```

You author source and notes locally. Droyd records durable workflow history,
versions, evaluations, submissions, and the views you see in the web app.

Start with [a workspace](/workflow/start-workspace), then use a small smoke
evaluation before moving to broader practice data.