OpenBrowse

Overview

OpenBrowse is the open source browser agent — a free, model-agnostic Chrome extension.

OpenBrowse is a free, open source Chrome extension that puts an AI agent in your browser. It can read pages, click, navigate, run code, and connect to external services via MCP — all under your control, with any model you choose.

Install

OpenBrowse is launching on the Chrome Web Store soon. While the listing is being approved, you can install the latest build manually in about 30 seconds.

Your data carries over

Manual installs share the same extension ID as the future Chrome Web Store install, so when you switch over later, your conversations, API keys, MCP connectors, memories, and workspaces will all persist automatically.

Manual install (current)

  1. Download openbrowse-<version>-chrome.zip (the latest version) from the latest GitHub Release.
  2. Unzip it anywhere on your machine. Keep the folder around — Chrome loads it from disk; deleting the folder uninstalls the extension.
  3. Open chrome://extensions and toggle Developer mode on (top right of the page).
  4. Click Load unpacked and select the unzipped folder.
  5. Pin OpenBrowse to your toolbar so you can find it easily — click the puzzle-piece icon in the top-right of Chrome, then the pin next to OpenBrowse.

That's it. No account, no signup.

Updating manual installs

Manual installs don't auto-update. To pick up a new release, download the new zip, replace the contents of the same folder, then click the refresh icon next to OpenBrowse in chrome://extensions.

Chrome Web Store (coming soon)

Once the listing is approved, install will be a one-click affair:

  1. Open the OpenBrowse listing on the Chrome Web Store. (Link goes live as soon as the listing is approved.)
  2. Click Add to Chrome, then confirm.
  3. If you have a manual install loaded, remove it first from chrome://extensions — Chrome won't run two extensions with the same ID simultaneously. Your data carries over to the Web Store install regardless.

What you can do

  • Browse and act — read page content, click elements, type, navigate, take screenshots
  • Organize tabs — group windows into Spaces with per-space color themes
  • Run quick actions — global command palette (⌥K) on every tab
  • Connect tools — MCP integrations for GitHub, Linear, Slack, Notion, and more
  • Stay private — run fully on-device with a local model; no server, no telemetry

Next steps

  • Comparison — how OpenBrowse compares to Claude for Chrome and others
  • Basic Usage — keyboard shortcuts and getting around
  • Agent — what the browser agent can do

On this page