Onevium
GitHubDownload
Core Workflows

Browser automation with Chrome

Set up the Claude in Chrome extension, connect it to Onevium, and let Claude navigate pages, fill forms, capture screenshots, and read console output.

Updated 2026-03-24

How browser automation works

Onevium connects to Chrome through a native messaging bridge. A Chrome extension communicates with the desktop app via a local socket — no network traffic leaves your machine. When the bridge is active, Claude gains access to 18 browser tools that it can use alongside file edits, terminal commands, and other capabilities.

Step 1: Install the Chrome extension

Install the "Claude in Chrome" extension from the Chrome Web Store. The extension is published by Anthropic and works with Chrome, Edge, Arc, Brave, Vivaldi, and other Chromium-based browsers.

Chrome Web Store link: https://chromewebstore.google.com/detail/claude-for-chrome/fcoeoabgfenejglbffodgkkbkcdhcgfn

Step 2: Register the native messaging host

After installing the extension, you need to register the native messaging host so the extension can communicate with Onevium. Run this command in your terminal:

claude --chrome

This creates a native messaging host manifest at the appropriate location for your browser and OS. Onevium detects this manifest automatically on startup.

  • macOS: ~/Library/Application Support/Google/Chrome/NativeMessagingHosts/com.anthropic.claude_code_browser_extension.json
  • Windows: Registry key under HKCU\SOFTWARE\Google\Chrome\NativeMessagingHosts\
  • Linux: ~/.config/google-chrome/NativeMessagingHosts/com.anthropic.claude_code_browser_extension.json

Step 3: Enable Chrome integration in Onevium

Open Onevium, go to the connectors panel, and enable Chrome integration. The status indicator shows the connection state — a green dot means the bridge is active.

The bridge communicates via a Unix socket on macOS/Linux (/tmp/claude-mcp-browser-bridge-{username}/) or a named pipe on Windows. No manual configuration is needed — discovery is automatic.

Available browser tools

Once connected, Claude has access to the following tools in every chat session. Claude chooses which tools to use based on your prompt, or you can ask it to use specific ones.

  • navigate — Open URLs, follow links, switch between tabs.
  • read_page — Extract page content, DOM structure, or specific elements by selector.
  • get_page_text — Get the full text content of the current page.
  • computer — Take screenshots and perform actions (click, type, scroll, drag).
  • form_input — Fill form fields by label or selector.
  • find — Locate elements on the page matching a query.
  • javascript_tool — Execute JavaScript in the browser console.
  • read_console_messages — Read console.log output, warnings, and errors.
  • read_network_requests — Inspect XHR/fetch requests and responses.
  • tabs_create_mcp — Open a new tab.
  • tabs_context_mcp — Get context about all open tabs.
  • resize_window — Resize the browser window to specific dimensions.
  • gif_creator — Record animated GIF captures of browser activity.
  • upload_image — Upload an image to the current page.
  • shortcuts_list / shortcuts_execute — List and trigger keyboard shortcuts.
  • switch_browser — Switch focus to a different browser.

Tab context and session isolation

Each Onevium chat session maintains its own browser tab context. When you start a new conversation, the bridge automatically discovers the active tab. If Claude opens new tabs during a session, they are tracked within that session's context.

If a tab is closed or becomes unavailable, the bridge auto-recovers by re-establishing context on the next tool call.

Common workflows

Browser automation works best for tasks that involve repetitive navigation, visual inspection, or data extraction from live pages.

  • QA smoke tests: "Open our staging site, log in with test credentials, run through the checkout flow, and screenshot any errors."
  • Competitive analysis: "Go to [competitor URL], capture their pricing page, and compare it to our current pricing."
  • Dashboard monitoring: "Open Grafana at [URL], screenshot the API latency panel, and summarize any anomalies."
  • Form testing: "Navigate to the signup form, fill it with edge-case data (empty fields, special characters), and report which validations fire."
  • Content auditing: "Check all links on our docs site and report any 404s."

Supported browsers

The native messaging host works with any Chromium-based browser. Onevium checks for manifests in the following locations:

  • Google Chrome (stable, Beta, Canary)
  • Microsoft Edge
  • Arc
  • Brave
  • Vivaldi
  • Opera
  • Chromium