SaaSActive

Browser Use

Browser Use Cloud offers V4, V3, and V2 agents that complete browser tasks.

Open source page ↗

Field note

What it does

Compare Browser Use Cloud agents on accuracy, speed and price: V4 is most accurate, V3 fastest, V2 legacy.

Capabilities

Available capabilities

Tags

Tags

Ways to use it

Ways to use it

sdk

python / browser-use

mcp

Not filed / https://docs.browser-use.com/

cli

Not filed / https://docs.browser-use.com/open-source/browser-use-cli

rest

Not filed / https://docs.browser-use.com/cloud/api-reference

pypi

python / browser-use-sdk

npm

typescript / browser-use-sdk

rest

Not filed / https://docs.browser-use.com/cloud/guides/billing

other

Not filed / https://docs.browser-use.com/cloud/guides/imessage

cli

Not filed / Not filed

other

Not filed / Not filed

sdk

python / browser-use-sdk

rest

Not filed / Not filed

rest

Not filed / https://docs.browser-use.com/cloud/api-v2-overview

pypi

python / browser-use-sdk

npm

typescript / browser-use-sdk

Product features

Product features

API Key Authentication

Use X-Browser-Use-API-Key header; do not put the API key in a Bearer header.

API-key monthly spending cap

Soft monthly USD cap per key; exceeding returns HTTP 402 api_key_monthly_spend_limit_reached.

Billing Account API

GET /api/v2/billing/account returns projectId, balance, and concurrentSessionLimit.

Billing account endpoint

GET /api/v2/billing/account returns totalCreditsBalanceUsd and components.

Browser API

Hosted browser without a hosted agent; connect through CDP.

Browser Infrastructure

Launch a cloud browser and connect to it from your code over CDP.

Browser lifetime

V4 browsers become eligible for cleanup after ~20 minutes idle and have a four-hour hard timeout.

Browser logins

Each linked chat gets its own workspace and, when enabled, a separate browser profile.

Stop Browser

PATCH /api/v4/browsers/{id} with {"action":"stop"} stops an unneeded managed browser.

Browser Use Agents

Give an agent a task and get the result.

browser-use-sdk

Hosted SDK with explicit v4 import for V4.

Browsers endpoint

POST /browsers returns id and cdpUrl; PATCH /browsers/{id} with stop action stops it.

CLI for your existing agent

Give Claude Code, Codex, Hermes, OpenClaw, or another agent the Browser Use CLI.

Cloud API V2

Lower-cost option for simple tasks.

Cloud API V4

Recommended API version for new agent integrations.

Connect a group

Get a number and code in Settings, add the number to a Messages group, and send the code to activate.

Create session

POST /api/v3/sessions to create a session with a task.

Fully hosted cloud

Send a prompt to the V4 agent API; Browser Use runs the agent and browser.

Get session result

GET /api/v3/sessions/SESSION_ID to retrieve session results.

Invoices

View and download invoices in Stripe under Settings → Billing → Billing History.

Live Preview and Recording

Open-source Python agent

Run and customize the Browser Use agent in your own code.

Open-Source Python Library

Separate open-source library with a different API from the hosted SDK.

Profile Sync

Copies selected cookies from local Chrome into a cloud profile; not automatic.

Project credits

Credits belong to a project; all keys in that project use its balance.

Runs endpoint

POST /runs creates a run; GET /runs/{id} returns the full run once terminal.

Session message queue

Queue up to 20 pending messages per conversation; a full queue returns 429.

Session queue endpoint

POST /sessions/{id}/queue queues a follow-up turn, or runs it immediately with interrupt: true.

Status endpoint

GET /runs/{id}/status is a cheap indexed lookup for polling until terminal.

Thinking levels

V2 model support, provider mappings, and a thinkingLevel request example.

Unlink a chat

You can unlink a chat in Settings when you no longer want it connected.

Usage tiers

Concurrency increases with net settled lifetime payments, from $0 (10 sessions) to $25,000 (1,000 sessions).

V2 Agent

Legacy first-generation agent, not actively maintained.

V3 Agent

Fastest; looks at the page and acts step-by-step like a human.

V4 Agent

Most accurate; writes and runs code for long, multi-step browser tasks.

Workspaces and Files

Runs have V4 workspace IDs; list generated files after the run finishes and request fresh download URLs.