Developers

Five minutes from signup to first session.

Grab an API key, install an SDK, point your existing agent framework at our runtime. Your first 60 session-minutes are on us, every month, forever.

1. Install

Pick your language.

Official SDKs for Python and TypeScript. Both wrap the Chrome DevTools Protocol, so anything that already speaks CDP works unchanged.

# Python
pip install browseranvil

# Node / TypeScript
npm install @browseranvil/sdk

# Curl (no SDK needed)
curl https://api.browseranvil.com/v1/sessions \
  -H "Authorization: Bearer ba_live_..." \
  -d '{"record": true}'
2. Start a session

One call. One browser.

The session object hands you back a Playwright-compatible browser or a raw CDP WebSocket URL — whichever your stack prefers.

from browseranvil import Runtime

runtime = Runtime(api_key="ba_live_...")

async with runtime.session(record=True) as s:
    page = await s.browser.new_page()
    await page.goto("https://example.com")
    await page.screenshot(path="hello.png")
3. Plug into your agent

Drop-in adapters for every major framework.

We maintain official adapters for Browser Use, Playwright, LangChain, and OpenAI Computer Use. One import change, agents migrate without touching logic.

# Browser Use
from browser_use import Agent
from browseranvil.adapters import browser_use_session

agent = Agent(
    task="Reconcile yesterday's invoices",
    browser=browser_use_session(record=True),
)

await agent.run()

Adapters, not rewrites.

The Switzerland of agent runtimes. Use us under whatever stack you already built on.

Browser Use

One-line --remote flag points the open-source library at BrowserAnvil.

Playwright

Standard connectOverCDP. Your existing tests and automations work as-is.

LangChain

Community BrowserAnvilToolkit plugs into any LangChain agent.

OpenAI Computer Use

Pipe the Computer Use tool calls into our runtime with the @browseranvil/cua package.

Anthropic Computer Use

Claude's Computer Use tool runs on our browser, with vaulted credentials included.

Puppeteer

Drop in a remote browser endpoint and keep your existing scripts.

LangGraph

Stateful agent graphs with our session lifecycle as a first-class node.

CrewAI / AutoGen

Sample integrations for multi-agent browser workflows in the community repo.

Built for the realities of agent code.

Things that usually take a week of infra work on day one of a new agent project — all defaults in the SDK.

  • Automatic retry and reconnect on transient network failures
  • Back-pressure and concurrency limits that respect your plan
  • Typed errors for bot-detection, auth failure, and navigation timeout
  • Streaming session events over WebSocket for live dashboards
  • First-class support for long-running sessions (hours, not seconds)
  • Telemetry hooks that slot into OpenTelemetry out of the box
API surface

A small, stable API.

POST /v1/sessions

Create a session. Returns a CDP WebSocket URL.

GET /v1/sessions/:id

Status, timing, and metadata for any session.

DELETE /v1/sessions/:id

Terminate early and release the microVM.

POST /v1/vault/credentials

Store or rotate a vaulted credential.

GET /v1/replays/:id

Fetch or stream a session recording.

POST /v1/webhooks

Register endpoints for session events and human-handoff triggers.

We had our first session running before the coffee got cold. Twenty minutes to a working prototype, including reading the docs.

— Kenji Tanaka, Founding Engineer, Kabuki Automations

Keep going.

GitHub

SDK source, examples, and adapter code. Pull requests welcome.

Reference docs

Every endpoint, every type, generated from the OpenAPI spec.

Changelog

Every SDK release and API change, in human language.

Status page

Live uptime per region. Subscribe for incident updates.

Discord

A few hundred agent builders who help each other debug.

Office hours

Weekly open Zoom with our engineers. Bring hard problems.

Your first agent session is one import away.

60 free minutes per month. No credit card. No trial clock.

Open the quickstart See pricing