Try a local example

See what Agentagon captures before using it on your application. This bundled example creates a tiny synthetic application, prepares an evidence packet, and saves a partial report. It needs no coding host, credentials, or network services after installation.

Run the example

Install the CLI, then run this from the Agentagon source checkout using the Python environment where it is installed:

python examples/local-audit/demo.py

Inspect the output

The example prints JSON with paths and IDs similar to these:

{
  "workspace": "/tmp/agentagon-example-…/app",
  "audit_id": "audit_…",
  "pending_action": "evidence",
  "packet": "/tmp/agentagon-example-…/app/.agentagon/audits/…/packets/….json",
  "response_template": "/tmp/agentagon-example-…/app/.agentagon/audits/…/packets/….response.json",
  "report": "/tmp/agentagon-example-…/app/.agentagon/reports/…/report.md"
}

Use the actual printed paths; the values above are abbreviated examples.

1. Open packet to see the captured evidence and review stage.
2. Open report to see coverage and pending work.
3. Notice that pending_action is evidence. The coding agent has not reviewed the evidence yet.

A prepared packet is not a completed audit

This example does not fabricate findings or claim the application passed. To obtain a real review, open the returned workspace in your coding host and ask Audit to resume its pending code audit.

View it in the dashboard

Substitute the printed workspace path:

agentagon --workspace /printed/workspace/path

The dashboard shows the unfinished audit. The shared local service remains available after the browser closes.

The temporary directory remains after the example exits. Preserve it if you want to continue; remove only that returned example directory when you are finished. The example uses its own settings file and does not change your application.

Next: Run your first audit on an application you want to improve.