AgentTrace

AgentTrace is a local-first command-line tool for recording audit trails of AI-assisted coding work.

It helps you answer:

What changed, what tests ran, what review happened, and what evidence supports the final result?

Start Here

If you are new to AgentTrace, read these first:

1. What AgentTrace Does
2. Install And Run
3. Your First Trace
4. Daily Workflow

Basic Workflow

init -> start -> snapshot -> add-test -> add-review -> report

In plain English:

1. Set up AgentTrace in your Git project.
2. Start tracking a coding task.
3. Capture the changed files and diff.
4. Record tests or manual QA.
5. Add review notes.
6. Generate a final report.

What AgentTrace Creates

AgentTrace writes local evidence into a .agenttrace folder in your project.

The most useful outputs are:

• report.md: the final report for one task
• pr-description.md: a pull-request-ready summary
• dashboard.html: a local dashboard across recorded runs