# For QA engineers

A QA engineer's path through EasyCode: acceptance criteria, the built-in browser, proof on the run and bugs filed as tasks.

## What does EasyCode do for QA?

EasyCode gives you and the agent the same browser. In a session in the desktop app, **Browser** opens a pane beside
the chat. You open the local app, a Storybook, or a **Preview or staging link** your team shared. The agent can drive
the same tab, check the page and save proof. You can record the page, mark parts of it with notes, and turn what you
found into a bug report on the task.

> **Desktop only**
>
> The built-in browser runs in the desktop app. On the web, a session shows **Open in the desktop app** or **Get the
> desktop app** instead. EasyCode staff can turn it off for an org with the [`browser`](/reference/kill-switches-and-modules#browser) switch.

## Your path

Follow these pages in order. Each line says why it matters to you.

### Get set up

1. [Quickstart](/setup/quickstart): join your org and open your first session.
2. [Desktop app](/setup/desktop-app): install it. The built-in browser and proof capture only run there.
3. [Harnesses on your own login](/setup/harnesses): the agent that drives the browser is your own harness.

### Know what to check

4. [From acceptance criteria to QA sign-off](/scenarios/acceptance-criteria-to-qa): how the criteria on a task become
   the checks you run and the proof you sign off on.
5. [Workflows, gates and the board](/guides/workflows-and-the-board): a gate in your workflow can wait for your
   **Approve** or **Request changes** before the run moves on.

### Test in the browser

6. [Test in the built-in browser](/guides/browser-testing): open pages, switch the device size, use the page tools,
   and choose a **Browser profile** from the tools menu.
7. [Test a staging link in the built-in browser](/walkthroughs/browser-qa): the same flow as a walkthrough of real
   screens.
8. [Pick up a session](/guides/pick-up-a-session): take over a developer's session to test their change with its
   full history.

### Prove it and report it

9. [Test in the built-in browser](/guides/browser-testing): the **Proof** panel shows each check, its recording and a
   **PR block** you can copy into the pull request.
10. [From idea to merged PR](/scenarios/idea-to-merged-pr): where your sign-off sits in the whole loop.

## What can the agent do in the browser?

The agent works through the browser tools: [`browser_navigate`](/reference/mcp-tools#browser-navigate) opens a page, [`browser_snapshot`](/reference/mcp-tools#browser-snapshot) reads it, and [`browser_click`](/reference/mcp-tools#browser-click) and [`browser_type`](/reference/mcp-tools#browser-type) act on it.
It never types into password, one-time-code or card fields; it asks you. [`proof_run`](/reference/mcp-tools#proof-run) runs a flow and
saves the proof. **What the agent did** in the tools menu lists its steps.

## How do I file a bug?

In the **Proof** panel, **Report a bug** builds a report from what happened: the steps to reproduce, the viewport,
the errors on the page, the requests that failed and your comments.

- **Copy report** puts it on your clipboard.
- **Send to agent** adds it to your message, so the agent can fix it.
- **Create bug** shows the bug first. **File bug** posts it as a sub-task of the task.

## Where does the proof go?

Proof is attached to the workflow run. In a private session it stays **On this computer**, and nothing is uploaded.
A pack marked **Held** stays on your computer until you look it over and choose **Attach anyway**.

## Where next

- [MCP tools](/reference/mcp-tools): every browser tool and what it returns.
- [Troubleshooting](/help/troubleshooting) when a page won't load in the pane.
