README
¶
Open Browser Use
open-browser-use is an open-source Chrome automation stack for AI agents. It
connects a Chrome MV3 extension, a Go native messaging host, and local SDKs so
agents can control a real Chrome profile through a readable, self-hosted route.
The project focuses on a small core: install the browser extension, install the local host, then use the CLI or SDKs to inspect tabs, open pages, run CDP commands, and hand browser state back to an upper-layer agent runtime.
Quick Start
Install the CLI
Install the native host and CLI with npm:
npm install -g open-browser-use
Or install with Homebrew:
brew install iFurySt/open-browser-use/open-browser-use
Set Up Chrome
Run open-browser-use by itself at any time to see the CLI version, whether the
browser extension was detected, the extension version when available, and the
next install or upgrade command.
After installing the CLI, register the native host and ask Chrome to install the Web Store extension:
open-browser-use setup
Restart Chrome and approve the Open Browser Use extension prompt if Chrome asks.
While the Chrome Web Store item is pending review, use the latest GitHub Release zip as an unpacked extension instead:
open-browser-use setup beta
setup beta downloads the latest
open-browser-use-chrome-extension-*.zip from
GitHub Releases
and uses its stable extension key to register the native host. It opens
chrome://extensions/ and reveals that same ZIP in Finder or the system file
manager. Enable Developer mode, then drag the ZIP file into the Chrome
extensions page to install it manually with the same id.
If you only need to repair the native messaging host registration, run:
open-browser-use install-manifest
Use It
Check that the extension and native host can talk to each other:
open-browser-use ping
List browser and session state:
open-browser-use info
open-browser-use tabs
open-browser-use user-tabs
Open and control a tab:
open-browser-use open-tab --url https://example.com
open-browser-use navigate --tab-id <tab-id> --url https://github.com/iFurySt/open-codex-browser-use
Run a small action plan when you want CLI-level orchestration without writing JavaScript or Python:
open-browser-use run -c '
name-session "Docs scan - OBU"
open-tab https://docs.browser-use.com
wait-load domcontentloaded
page-info
finalize-tabs []
'
Each line is a supported browser action. open-tab and claim-tab set the
default tab for later tab-scoped actions.
The CLI also exposes lower-level Browser Use style calls through:
open-browser-use call --method <method> --params '<json>'
Directories