> ## Documentation Index
> Fetch the complete documentation index at: https://docs.hud.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Browser and computer use

> Build and evaluate tasks where agents operate a real browser or desktop, with a screenshot at every step and QA agents that read them.

For tasks where the agent works the way a person does: in a real browser, or on a desktop with a
screen, keyboard, and mouse. HUD hosts the browser or desktop for each run, records a screenshot at
every step, and QA agents read the screenshots along with the rest of the trace.

## Start from

| Starting point | What it is |
| - | - |
| **browser-use** | A featured environment. The agent researches a rental listing in real Chromium, and its answer must cite the screenshots it relied on. Run its `rental-kitchen-chairs` template from the web app with no setup. See [featured environments](/platform/create-tasks#featured-environments). |
| **Computer Use** | An environment template with a full desktop. Start from it with `hud init my-env --template cua`. |
| **Browser** and **Deep Research** | Environment templates in the **Agentic Collaboration** group. On [Environments](https://hud.ai/environments), select **New Environment**, then **See all templates**. |

## How it works

An environment exposes a browser over the Chrome DevTools Protocol, a desktop screen over VNC, or
both. HUD's built-in agents open the ones they support and give the model its provider's tools for
them, so tasks don't change from one model to the next. See
[capabilities](/v6/reference/capabilities).

<Steps>
  <Step title="Deploy the environment">
    Deploy your environment with `hud deploy`, or use a featured one. See
    [environments](/platform/environments).
  </Step>

  <Step title="Create tasks">
    Each template defines a kind of task; each task fills in its arguments, such as the site to
    start on and what to find. See [create tasks](/platform/create-tasks).
  </Step>

  <Step title="Run them on several models">
    Start a job from the taskset and compare models on the same tasks. See
    [evaluations](/platform/evaluations). A trace holds at most 500 screenshots, so keep
    **Max steps** at 500 or below when the agent takes one per step. See
    [trace limits](/platform/limits#traces).
  </Step>

  <Step title="Watch the runs">
    A trace's **Screen** view shows the screenshots from its browser and computer-use steps, and
    the conversation shows the action behind each one. See
    [inspect a trace](/platform/traces#inspect-a-trace).
  </Step>

  <Step title="Review them with QA">
    QA agents view each screenshot, so they catch answers that the screens don't support and
    failures caused by the page rather than the agent. See [QA agents](/platform/qa-agents).
  </Step>
</Steps>

## Grade what the agent achieved

Grade the outcome, not the clicks. Have the template ask for a
[structured answer](/v6/reference/types#typed-task-io) and check it against the source, or check the
state the agent left behind, such as a submitted form or a saved file. The browser-use template
gives full reward only when the chair count is right and the screenshots the answer cites show the
kitchen, so a correct guess without evidence scores nothing.

<Columns cols={3}>
  <Card title="Traces" icon="https://mintcdn.com/hud-f5fd7c15/9ozVUWR-rlBrAGzN/platform/icons/trace.svg?fit=max&auto=format&n=9ozVUWR-rlBrAGzN&q=85&s=e3c55aebd34125404cef2f9def8baa6e" href="/platform/traces" width="24" height="24" data-path="platform/icons/trace.svg">
    Replay every step and screenshot.
  </Card>

  <Card title="Capabilities" icon="plug" href="/v6/reference/capabilities">
    Browser, screen, shell, and tool protocols.
  </Card>

  <Card title="Run QA" icon="play" href="/platform/run-qa">
    Review one trace or every new run.
  </Card>
</Columns>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.