# Browser-host compatibility and test evidence

Check where Aluvia can configure Chrome, which runtime checks have evidence, and what still needs verification in your agent environment.

Aluvia is a local CLI and proxy daemon. It needs command access to the machine running a usable Chrome or Chromium browser and Node.js 18 or later. A browser on an inaccessible remote service cannot be configured from a separate agent container.

The table separates recorded tests from setup guidance. It does not certify every agent product or establish access to any target website.

## Recorded verification

| Runtime or use case | Evidence and date | What the evidence covers |
| --- | --- | --- |
| Debian 12, Node 22.23.2, Chromium 152.0.7977.82 on Xvfb | 2026-09-07: [packed CLI browser smoke report](https://github.com/aluvia-connect/aluvia/blob/99a81d901fad913318414e1f25976f5475d92948/docs/testing/bare-setup.md) | CLI 1.4.12 configured real Chromium, kept the tested profile/cookie and debugging port, installed the skill and reused the session on repeat setup. Network services were local fixtures, not a production trial. |
| Node 18.20.8, 20.20.2 and 22.23.2 | 2026-09-07: [automated runtime checks](https://github.com/aluvia-connect/aluvia/blob/99a81d901fad913318414e1f25976f5475d92948/docs/testing/bare-setup.md#automated-checks) | 202 automated tests passed per runtime; three Windows-only tests skipped. These are not browser tests on every operating system. |
| macOS Chrome | 2026-09-07: [recorded test limits](https://github.com/aluvia-connect/aluvia/blob/99a81d901fad913318414e1f25976f5475d92948/docs/testing/bare-setup.md#limits) | Documented product platform; no end-to-end macOS browser restart evidence in this report. Check setup output and the actual browser task. |
| Grok Bot, Codex, Cursor, Claude, OpenClaw and Hermes | 2026-09-08: [host-specific guides](/geo) | Conditional setup guidance, not end-to-end integration certification. The agent needs a shell on its browser host; model availability and built-in search routing are separate. |
| Playwright | 2026-09-08: [browser/context prerequisites](/geo/playwright) | Check the specific browser and context proxy configuration. The Linux smoke test does not certify every Playwright launch, remote browser or pipe-based controller. |

## Inspect the skill before setup

Read the [current source skill](https://github.com/aluvia-connect/aluvia/blob/main/skills/aluvia/SKILL.md) and the [skill bundled with published CLI 1.4.12](https://github.com/aluvia-connect/aluvia/blob/99a81d901fad913318414e1f25976f5475d92948/packages/cli/skills/aluvia/SKILL.md). A source update is not a package release. Setup installs the copy bundled with the version you run.

## Try an eligible browser task

First 10 MB free without an account, API key or payment; then $2/GB. Initial setup can restart Chrome. Follow the [quickstart](/docs): inspect setup, select an available country if needed, reload the page and check the actual content. `ready: true` confirms the proxy path, not task success.

---

Source: https://aluvia.io/compatibility