OpenClaw with Playwright or Puppeteer: how to install browser support

If you want the broader introduction first, read What is OpenClaw and how can it help?. This post is narrower: it focuses on browser automation, screenshots, and what to install if you want that part to work reliably.

The short answer is:

• if you want browser automation inside OpenClaw, treat Playwright as the supported path;
• if you were searching for Puppeteer + OpenClaw, you are probably really looking for browser control and screenshots, and OpenClaw’s official docs point to Playwright for that;
• if your workflow is screenshot only, using ScreenshotOne can be simpler than running browser infrastructure inside your agent stack.

What OpenClaw can do with the browser

As of March 12, 2026, the official OpenClaw docs describe a browser tool that gives your Claw an isolated browser profile and lets it:

• open tabs and navigate pages;
• read page content and make AI-readable snapshots;
• click, type, select, drag, and scroll;
• take screenshots of the full page or specific elements;
• save PDFs;
• work with a local browser or a remote CDP browser.

That matters because screenshots are only one part of the story. OpenClaw can use the browser to inspect a site, log in, click through a flow, verify the result, and then send back a screenshot as evidence.

So the browser is useful for:

• monitoring dashboards or status pages;
• checking competitor pages or pricing pages;
• opening internal tools and confirming state changes;
• collecting visual proof after a workflow runs;
• building chat-first assistants that can browse and report back.

Playwright vs Puppeteer in OpenClaw

This is the part that is easy to misunderstand.

OpenClaw does not position Puppeteer as the main browser layer in its docs. The browser documentation instead describes a CDP-backed setup with Playwright used for advanced browser actions.

The docs also explicitly mention a Playwright requirement:

• if Playwright is missing from the Gateway build, advanced actions return a 501 error;
• OpenClaw recommends installing the full playwright package, not playwright-core;
• or reinstalling OpenClaw with browser support enabled.

So if your goal is “make browser tasks and screenshots work in OpenClaw,” the practical answer is:

• use OpenClaw’s browser tool;
• make sure the deployment includes Playwright;
• treat Puppeteer as a separate browser automation preference, not as the documented OpenClaw integration path.

That does not mean Puppeteer is a bad tool. It just means that if you want the least-friction setup inside OpenClaw, follow the Playwright path the product documents today.

Install OpenClaw first

If you are starting from scratch, install OpenClaw with the official installer:

1
curl -fsSL https://install.openclaw.ai | bash

Then run onboarding and verify the Gateway:

1
openclaw
2
openclaw doctor
3
openclaw status
4
openclaw dashboard

The browser feature is controlled by the Gateway configuration. The docs show it under browser.enabled, and the browser profile is intentionally separate from your personal daily browser.

Install the browser support the right way

Standard local install

If you installed OpenClaw through the recommended installer, browser support may already be present. If browser actions fail with a message that Playwright is unavailable, the docs point you to the fix:

• install the full playwright package in the same environment as the Gateway;
• do not use playwright-core for this;
• restart the Gateway after installation;
• if needed, reinstall OpenClaw with browser support.

For a Node-based install, that usually means:

1
npm install playwright

The important detail is not the package manager syntax. The important detail is that OpenClaw expects the full Playwright package in the Gateway environment.

Docker install

The docs are more explicit for Docker-based deployments. They show this command for adding Playwright support inside the OpenClaw container:

1
docker exec openclaw npm install playwright

Then restart the container or Gateway so OpenClaw can pick it up.

How to make screenshots in OpenClaw

Once browser support is working, you have a few options.

1. Use the browser CLI

OpenClaw documents browser commands like these:

1
openclaw browser --browser-profile openclaw start
2
openclaw browser --browser-profile openclaw open https://example.com
3
openclaw browser --browser-profile openclaw screenshot --full-page

That is the simplest way to prove the browser is alive and can capture a page.

If you want an element-level screenshot, first create a page snapshot so you can reference a specific node, then capture that element:

1
openclaw browser --browser-profile openclaw snapshot
2
openclaw browser --browser-profile openclaw screenshot --ref "<element-ref>"

The same tool can also generate PDFs:

1
openclaw browser --browser-profile openclaw pdf

2. Use the OpenClaw UI

The docs also show a browser control UI in the dashboard. That is useful when you want to:

• inspect the current page manually;
• see what the browser session is doing;
• log in to sites that are easier to authenticate interactively;
• test a screenshot flow before turning it into an automated skill.

3. Use it from the agent

This is where OpenClaw gets interesting. Because the browser is part of the assistant environment, you can ask the Claw to:

• open a website;
• check whether something changed;
• click through a flow;
• take a screenshot;
• send the screenshot back to you in chat.

That is much more useful than a raw screenshot script when the screenshot is just the final output of a broader workflow.

Where Puppeteer still fits

If you personally prefer Puppeteer, you can still use it in your own application stack outside OpenClaw. But that is different from asking, “What should I install so OpenClaw browser actions work?”

For that question, the official answer is effectively Playwright.

So the practical split is:

• use OpenClaw + Playwright when the screenshot is one step inside a bigger agent workflow;
• use Puppeteer if you are building your own browser automation separately from OpenClaw;
• use ScreenshotOne if you only need reliable screenshots and do not want to manage browsers at all.

Why ScreenshotOne can be the better choice for screenshot-only flows

OpenClaw’s browser is valuable when you want a real assistant that can browse, click, think, and report back.

But if the whole job is just:

1. open a URL;
2. render it correctly;
3. return a clean screenshot;

then running a browser stack inside your agent platform can be more moving parts than you need.

That is where ScreenshotOne has a clear advantage.

With ScreenshotOne, you do not need to worry about:

• browser lifecycle management;
• Playwright provisioning;
• long-lived browser sessions;
• headless browser crashes or memory spikes;
• tuning screenshot rendering edge cases by hand.

And because ScreenshotOne is focused specifically on screenshots, you get screenshot-first features that are often useful in production:

• blocking cookie banners;
• blocking chat widgets;
• consistent capture at scale;
• simple API-based integration;
• cleaner output for reports, monitoring, or archival.

So the decision is not “which product is better?” It is “where is the complexity worth carrying?”

• If you need a browser-capable assistant, OpenClaw is the right layer.
• If you need a screenshot service, ScreenshotOne is the simpler layer.

In many teams, the best setup is both:

• let OpenClaw decide when and why to capture something;
• let ScreenshotOne handle rendering the screenshot itself.

If that is your use case, also read Screenshots for OpenClaw Workflows.

Summary

If you searched for how to install Puppeteer or Playwright with OpenClaw, the main thing to know is that OpenClaw documents Playwright, not Puppeteer, as the browser dependency that matters for advanced actions.

Use OpenClaw’s browser when you want the assistant to navigate, click, inspect, and verify. Use ScreenshotOne when the workflow is really about screenshot delivery, not browser orchestration.

That split will save you time and keep your setup simpler.

Sources

• OpenClaw website
• OpenClaw installation docs
• OpenClaw browser docs

Frequently Asked Questions

If you read the article, but still have questions. Please, check the most frequently asked. And if you still have questions, feel free reach out at support@screenshotone.com.

Does OpenClaw use Puppeteer or Playwright?

As of March 12, 2026, OpenClaw's browser docs describe a CDP-backed browser layer with Playwright used for advanced actions. The docs do not describe Puppeteer as the primary integration path.

Can OpenClaw take screenshots of websites?

Yes. OpenClaw's browser tool can take page screenshots, element screenshots, and PDFs, either through the UI, the CLI, or agent workflows.