iFurySt/open-browser-use

Open Browser Use – A Neutral Browser‑Automation Layer for AI Agents

What it is – open-browser-use is an open‑source tool that lets AI agents (e.g., Codex, Claude‑Code, or any custom runtime) control a web browser. It consists of a small Chrome extension paired with a command‑line interface (CLI) and language‑specific SDKs (JavaScript/TypeScript, Python, Go). The design purpose is to stay runtime‑agnostic: the same browser‑control capabilities can be used from any agent that can talk to the CLI or the MCP (Message‑Control‑Protocol) server.

Why it matters – Large‑language‑model agents often need up‑to‑date web information (news, documentation, search results). Codex.app recently introduced a proprietary Browser Use capability; open-browser-use provides a free, community‑maintained alternative that works with multiple runtimes and can be installed locally.

---

Key Features (as described in the README)

• Cross‑runtime neutrality – works with any agent that can invoke a CLI or MCP server.
• Browser extension + native host – the extension talks to a native host registered by the open-browser-use CLI, enabling low‑level Chrome DevTools Protocol (CDP) commands.
• SDKs for three major languages –
  • JavaScript/TypeScript (open-browser-use-sdk on npm)
  • Python (open_browser_use on PyPI)
  • Go (obu package)
• MCP server – exposes a set of browser tools (list tabs, open/claim tabs, navigate, run CDP actions, clean‑up) via a simple obu mcp command that agents can call.
• Skill packages – ready‑to‑install .skill files for Codex and Claude‑Code, plus a generic skill directory in the repo.
• Easy installation – global npm package, Homebrew tap, or direct binary download; a setup command registers the native host and opens the Chrome Web Store page for the extension.
• Manual ZIP fallback – if the Web Store is unavailable you can install the extension from a ZIP you provide.

---

Quick‑Start Summary

1. Install the CLI (global npm or Homebrew):

   npm i -g open-browser-use   # or brew tap iFurySt/open-browser-use && brew install open-browser-use
   

2. Run the setup wizard – registers the native host and opens the Chrome Web Store page for the companion extension:

   open-browser-use setup
   

   If the store is down, use open-browser-use setup beta or point to a local ZIP.
3. Add an SDK in the language you’ll use:

   # JS/TS
   npm install open-browser-use-sdk
   # Python
   pip install open-browser-use-sdk
   # Go
   go get github.com/ifuryst/open-browser-use/packages/open-browser-use-go
   

4. Install the skill for your agent runtime (example for Codex):

   npx skills add iFurySt/open-browser-use -g -a codex --skill open-browser-use -y
   

5. Start the MCP server (optional, for direct tab control):

   npx add-mcp "obu mcp" --name open_browser_use --all -g -y
   

   Agents can then call the MCP methods to list tabs, open URLs, navigate, etc.

---

Typical Use Cases

• Web‑search augmentation – an LLM agent can fetch the latest Hacker News headlines, read the article, and produce a summary.
• Data extraction – navigate to a dashboard, scrape tables via CDP, and feed the result back to the model.
• Automated testing / UI interaction – agents can drive a browser to perform repetitive UI tasks.
• Multi‑agent orchestration – because the tool is runtime‑neutral, different agents (Codex, Claude‑Code, custom bots) can share the same browser control service.

---

License

The project is released under the permissive MIT License.

---

Where to Learn More

• Deep Dive blog post – https://www.ifuryst.com/en/blog/2026/open-browser-use/
• GitHub releases – pre‑built binaries, skill packages, and extension ZIPs.
• DeepWiki badge – points to additional community documentation.