Docs·Quickstart

Quickstart

Connect a GitHub repo, hand your coding agent an API key, and Omea verifies real behavior in a clean cloud environment — streaming verdicts your agent can act on and leaving evidence you can open. About five minutes to your first run.

1. Connect your repository

   Sign in and connect a GitHub repository through Omea’s read-only GitHub App. Omea creates an application for that repo and starts a first verification run on its own — no configuration, no changes to your code.

2. Create an API key

   Open your app’s Overview and, under Connect your coding agent, create an API key. It’s scoped to this one application, never expires, and you can revoke it anytime. Copy it once — the full key is shown only at creation.

3. Give the key to your agent

   Pick whichever fits your setup — there’s no separate login step, and CI environments that can only pass instructions still work:

   # store it once on your machine $ npx @omea/cli login --key omea_xxx # or set it in the environment (CI) $ export OMEA_API_KEY=omea_xxx # or pass it inline on any command — zero pre-auth $ npx @omea/cli test --api-key omea_xxx

4. Run the existing suite

   $ npx @omea/cli test

   Omea runs your repository’s durable test suite in a clean, isolated cloud VM — a real browser exercising the real product — streams each pass/fail verdict as it completes, computes coverage, and lands the run in your dashboard. It never edits your app or your tests.

5. Grow coverage

   $ npx @omea/cli grow "cover the checkout and refund flows"

   Omea’s agent writes new end-to-end tests toward the flows you name — or the least-covered code if you leave the message off — runs them, and keeps the ones that hold. Kept tests join your durable suite, so every future omea test runs them too.

6. Read the evidence

   Every run lands at omeahq.com/dashboard/<org>/apps/<app>/runs/<run>: the verdict, coverage, the agent’s full conversation, video and trace evidence, and any findings a human should look at. Trust isn’t a green checkmark from the agent that wrote the change — it’s proof you can reopen.

Point a coding agent at Omea

# Omea — runtime verification
Omea verifies the behavior your app already has.

Run the suite       →  omea test
Grow coverage       →  omea grow "focus area"

Each returns the verdict for the agent to act on.
The output also links a human to the video evidence.

Commands

What you don’t have to do

No test harness to wire up, no environment to reproduce, no secrets pasted into a config file, and no edits to your repository — connecting it and holding an API key is the whole setup. Omea boots the app, provisions what it needs, and runs against the real thing.