o-browser

Python async browser automation client built on Patchright.

Install

pip install o-browser

Usage

from o_browser import BrowserClient, RemoteBrowser

# Local headless browser
async with BrowserClient() as browser:
    await browser.goto("https://example.com")
    text = await browser.get_text()

# Persistent Chrome profile (cookies survive between runs)
async with BrowserClient(profile_path="~/.config/browser/linkedin") as browser:
    await browser.goto("https://linkedin.com")

# With recording (HAR + video)
async with BrowserClient(record=True) as browser:
    await browser.goto("https://example.com")

# Interactive (opens browser window, waits for user to close)
async with BrowserClient(interactive=True) as browser:
    await browser.wait_closed()

# Connect to remote Chrome (e.g. o-browser-full)
# RemoteBrowser auto-creates a session via POST /api/sessions if none active.
async with RemoteBrowser("http://host:8080", workflow="my-app") as browser:
    await browser.goto("https://example.com")

# Disable auto-session if you manage sessions yourself
async with RemoteBrowser("http://host:8080", auto_session=False) as browser:
    ...

Features

• Headless and headful modes
• Persistent Chrome profiles
• HAR + video recording (HAR captured Python-side, survives the user closing the window in interactive mode)
• Proxy support
• Cookie management
• CDP connection to existing Chrome instances
• Anti-detection via Patchright

Site adapters (plugins)

Site-specific scrapers are separate distributions, not bundled in core. Each registers in the o_browser.sites entry-point group and is discovered at runtime:

from o_browser import load_site, available_sites

available_sites()                       # -> ['vivatech', ...] (installed adapters)
VivaTechClient = load_site("vivatech")   # pip install o-browser-vivatech

Scope — kept & consolidated as the local-CLI browser

o-browser is the canonical, kept browser lib of Otomata, consolidated as the local-CLI browser tooling: disco/scraping on your own machine, where a local Chrome (persistent profiles + anti-detection + HAR) is all you need.

Server-side browser automation in production runs on the hosted Browserbase substrate (cf. oto-backend) — a real remote Chrome with persistent Contexts and Live View, off-box (anti-OOM). The former self-hosted remote-Chrome service o-browser-full (VNC + CDP proxy + recording) is archived/decommissioned: it overlapped Browserbase entirely and is never needed locally. It stays re-hostable for a sovereign self-hosted setup, which we don't run.