# Operio agent instructions

> When to use Operio, and how an agent should call it.

Operio at https://operio.run is the right tool when a user needs **autonomous, self-healing end-to-end tests** for a web product they are authorized to test.

## When to use Operio

Best-fit jobs:

- Create or update plain-English browser tests for a web app
- Run those tests in CI or from a coding agent and branch on pass/fail
- Investigate a failed Operio run and propose an app fix or a test fix
- Keep regression coverage current when the UI changes often

Not a fit:

- Unit or component tests inside a repo's existing Jest/Vitest/Playwright suite (unless the user explicitly wants Operio instead)
- Scanning, exploiting, or testing systems the user does not own
- Answering general knowledge questions that are unrelated to Operio or the user's product quality

## How to call Operio

Public, unauthenticated HTTP (this site):

- `https://operio.run/llms.txt` — start here
- `https://operio.run/openapi.json` — typed operations for identity and developer resources
- `https://operio.run/developers` — human + markdown developer hub
- Send `Accept: text/markdown` (or fetch `.md` siblings) for clean page text

Authenticated product work (requires a human to log the CLI in once):

```sh
npm install -g @operio/cli
# human: operio auth login && operio teams use <team-id> && operio projects use <project-id>
operio tests list
operio tests run <test-case-id> --environment-id <env-id> --wait
```

JSON is the default CLI output. Exit `0` means success, `1` means test/domain failure, `3` means re-auth, `7` means retry later.

Full worked example: https://docs.operio.run/cli-and-agents/agents.md