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

> ## Agent Instructions
> Anchor Browser provides cloud browsers for AI agents and automation: stealth browsing with residential proxies, managed authentication into third-party web apps, and browser sessions that run reliably at scale. AI agents that need Anchor credentials should start with https://docs.anchorbrowser.io/quickstart/agent-access. This flow lets agents obtain an API key programmatically without creating a dashboard account. Read that page before attempting authentication or API access.

# Sandboxes

> Run agent or automation code in a third-party sandbox and connect it to an Anchor cloud browser.

A sandbox is where your agent or automation **code** runs. Anchor is where the **browser** runs.

The integration Anchor is shipping now is **Anchor SDK in the sandbox base image → Anchor Cloud**. The sandbox stays vanilla and lightweight. Chromium runs as a managed Anchor cloud session; the sandbox just holds the SDK and talks to it over HTTPS + CDP.

Keep those roles separate:

* The sandbox stays the code runtime — install packages, run a script, call a model.
* Anchor stays the remote browser — stealth, proxies, live view, and session lifecycle.
* Connect the two over HTTPS + CDP. Do not launch Chromium inside the sandbox.

This is the same pattern as [Stagehand](/integrations/stagehand): create an Anchor session, then point your tool at `cdp_url` when you need full Playwright control.

<Note>
  Anchor's own [Execute Code Snippet](/tools/execute-code-snippet) tool is different — it runs TypeScript *inside* an existing Anchor session. Pages in this section attach a sandbox *from the outside*.
</Note>

## What Anchor provides

* SDKs on [npm](https://www.npmjs.com/package/anchorbrowser) and [PyPI](https://pypi.org/project/anchorbrowser/) (Playwright or Puppeteer via CDP)
* Co-branded docs template (MDX)
* An Agent Skill (`SKILL.md`) so agents use the tool correctly — bundle it into the partner's default agent environment or skills catalog
* [Agent Access](/quickstart/agent-access) — a limited keyless trial, not a BYOK replacement

## Shared setup

Every partner page below uses the same technical contract:

* Preinstall `anchorbrowser` in the image (and `playwright` / `playwright-core` only if you need direct CDP control).
* Alias `ANCHORBROWSER_API_KEY=${ANCHOR_API_KEY}` — the raw SDK reads `ANCHORBROWSER_API_KEY`. Without that alias, session and task calls fail auth (401) even when `ANCHOR_API_KEY` is set.
* Customers bring their own Anchor API key (BYOK). Anchor bills them directly on usage.
* Prefer `agent.task` / `agentTask` for declarative browser work. The result string is `result.data.result`.
* Never log `session.data.cdp_url` — it carries the API key in the query string. Surface `session.data.live_view_url` to humans instead.
* Reuse one session per agent run, set `timeout.max_duration` / `timeout.idle_timeout`, and delete the session on sandbox teardown.

## Egress

If the sandbox restricts egress, allowlist:

* `https://api.anchorbrowser.io`
* `wss://connect.anchorbrowser.io`
* `https://live.anchorbrowser.io`

<CardGroup cols={2}>
  <Card title="CreateOS" icon="box" href="/integrations/sandboxes/createos">
    Isolated Linux VMs for agent code, connected to an Anchor cloud browser.
  </Card>

  <Card title="Daytona" icon="box" href="/integrations/sandboxes/daytona">
    Daytona sandboxes as the code runtime, with Anchor as the remote browser.
  </Card>

  <Card title="Tenki" icon="box" href="/integrations/sandboxes/tenki">
    Disposable Linux VMs for agents, connected to an Anchor cloud browser.
  </Card>
</CardGroup>