Browser Control & Canvas
Automate web browsing and use the visual Canvas workspace with OpenClaw
Automate web browsing and use the visual Canvas workspace with OpenClaw
OpenClaw includes powerful browser automation capabilities and a visual Canvas workspace. These features allow the agent to interact with websites, automate tasks, and provide a visual interface for complex operations.
Built-in browser automation works everywhere but can be slow and brittle: the agent launches Chrome, waits for render, and scrapes the DOM. Many sites already expose the same data and actions via internal APIs. Unbrowse is an OpenClaw plugin that captures those APIs once (via network interception) and turns them into direct API calls your agent can use. First visit uses the browser; every visit after is ~200 ms instead of 10–45 seconds. Good for repeated actions on the same site (trading, forms, monitoring). See the Unbrowse guide for how it works and when to use it vs the built-in browser.
PinchTab is a standalone HTTP server that gives AI agents direct control over Chrome. ~800 tokens per page with text extraction (5–13x cheaper than screenshots), multi-instance support (parallel Chrome processes with isolated profiles), and a 12MB Go binary. Run it as a separate service; the OpenClaw pinchtab skill lets your agent navigate, snapshot, click, fill, and extract via HTTP. Good for token-conscious automation, headless or headed mode, and Raspberry Pi (ARM64). Install: curl -fsSL https://pinchtab.com/install.sh | bash or npm install -g pinchtab. See pinchtab.com/docs. For deploying OpenClaw (including browser automation) on a remote server, see our VPS hosting guide.
Some sites—X.com (Twitter), LinkedIn, Product Hunt, Amazon, and many protected or paywalled blogs—block standard scrapers and headless browsers with 403 Forbidden or Cloudflare “Are you a human?” challenges. When the built-in browser hits these walls, the agent can’t get the content. On a VPS or remote server you’re typically limited to headless Chrome or raw HTTP, both of which are commonly blocked. See Protected Sites for a full guide. Two complementary options:
openclaw plugins install @askjo/camofox-browser) that runs a headless browser server built on Camoufox (Firefox fork, C++ fingerprint spoofing). Exposes tools like camofox_create_tab, camofox_snapshot, camofox_click, camofox_navigate. Best for interactive browsing on X, Product Hunt, Amazon from a server.This site does not endorse or guarantee third-party plugins; see the Skills list and each project’s repo for details.
The browser tool enables OpenClaw to:
Enable browser control in your configuration:
{
"browser": {
"enabled": true,
"color": "#FF4500"
}
}Chrome extension relay (2026.2.19+): The relay requires gateway-token auth on both /extension and /cdp. Use a single gateway.auth.token in the Chrome extension setup for relay authentication. If the relay port is already in use by another OpenClaw process, the Gateway can reuse that relay instead of failing; it still fails on non-relay port collisions.
Canvas is OpenClaw's visual workspace feature:
The Canvas host serves visual interfaces:
/__openclaw__/canvas/ pathsAsk OpenClaw to fill and submit forms:
Extract information from websites:
Monitor websites for changes:
Browser control is powerful and should be used carefully:
For untrusted sessions (groups), browser access is typically disabled by default.
Canvas provides a visual interface where:
Canvas is accessible via:
Common browser issues and solutions: