Integration

Browser Use integration

Browser Use Cloud runs AI agents that complete tasks in stealth cloud browsers. Start an agent run on a natural-language task and get its text result or structured JSON output, follow up in the same conversation, read run events and costs, share a session, manage workspaces (persistent files) and their files, browser profiles (saved logins) and Chrome extensions, launch standalone cloud browsers to drive over CDP, list their downloads, and read the account's credit balance and concurrency limit. This is the V4 API (the vendor's recommended version; V2 and V3 are not exposed). Uploading workspace files (presigned URLs), connected-app OAuth management and pay-with-crypto (x402) routes are not exposed. Authentication is an API key (starts with bu_) sent in the X-Browser-Use-API-Key header, without a Bearer prefix.

Connect Browser Use How integrations work

Ask it in chat

Say what should happen.
The workflow builds itself.

When this happens, do that in Browser Use.Built from the 45 Browser Use tools below, on the connection you add once.

What you can do with Browser Use

Each of these is one operation on one endpoint, available both as a workflow step and as a tool an AI agent can call on its own. Nothing bundles several operations into one action, so what a run did stays readable afterwards.

Most endpoints run on the free plan. A few, mainly the ones that publish on your behalf, need a paid plan.

Profile, 5 operations

create_profileCreate profilePOSTCreates a browser profile, which keeps cookies, local storage and login state between browsers. Use one per end user and pass its id as profileId in browser settings.
delete_profileDelete profileDELETEPermanently deletes a browser profile and its saved browser state.
get_profileGet profileGETReturns one browser profile: name, your userId, the domains it holds cookies for, and when it was last used.
list_profilesList profilesGETLists the project's browser profiles (saved cookies and login state) with their names, your userId and cookie domains; query searches by name or userId.
update_profileUpdate profilePATCHUpdates a browser profile's name or userId.

Browser, 4 operations

create_browserCreate browserPOSTStarts a standalone stealth cloud browser and returns its CDP URL (to drive it with Playwright or Puppeteer; Selenium cannot attach to it) and live view URL. Billed per hour upfront; stop it with stop_browser.
get_browserGet browserGETReturns one cloud browser: status, live view and CDP URLs, timeout, costs, labels, extensions and, once it has stopped, the presigned recording URL when recording was on.
list_browsersList browsersGETLists the project's cloud browsers with status, live view and CDP URLs, costs and labels; filter by status, agent session or metadata label. Recording URLs are only in get_browser.
stop_browserStop browserPATCHStops a cloud browser and returns it. Unused time is refunded (billing is per minute, minimum 1 minute).

Extension, 4 operations

delete_extensionDelete extensionDELETEPermanently deletes an uploaded extension. Browsers already running with it are not affected.
get_extensionGet extensionGETReturns one uploaded Chrome extension: its name, manifest version and upload date.
list_extensionsList extensionsGETLists the Chrome extensions uploaded to the project (name, version, upload date), one page at a time.
upload_extensionUpload extensionPOSTUploads a Chrome extension (a Manifest V3 ZIP with manifest.json at its root) to the project. Pass its id in extensionIds when creating a browser or a run.

Queue, 4 operations

cancel_session_messageCancel session messageDELETERemoves a queued message that is still pending and returns it. Fails with 409 once the session has picked the message up.
get_session_messageGet session messageGETReads one queued message of a session: its status (pending, dispatching, consumed, cancelled, superseded, failed) and the runId it started once dispatched.
list_session_queueList session queueGETLists a session's open queued messages, its recently delivered ones and its steering cutoffs (the earliest accepted interrupt of each interrupted run).
send_session_messageSend session messagePOSTSends a follow-up instruction to an agent session. If the session is busy the message waits in its queue and runs as the next turn; if idle it runs at once.

Session, 4 operations

delete_sessionDelete sessionDELETEDeletes an agent session: it disappears from lists and can no longer be opened or continued. An active run and pending queued messages are cancelled first. Records are kept for billing.
get_sessionGet sessionGETReturns one agent session: its workspace, latest run id, opening task, title and the latest run's status (use it to see whether the conversation is busy or idle).
list_sessionsList sessionsGETLists the project's agent sessions (conversations), one row per session showing its latest run's task, title and status, most recent first. Pages with cursor.
rename_sessionRename sessionPATCHRenames an agent session (title up to 255 characters). A null or blank title clears the name and the dashboard shows the opening task instead.

Workspace, 4 operations

archive_workspaceArchive workspaceDELETEArchives a workspace. Idempotent for missing, already archived and foreign ids. Archiving is not a file deletion: use delete_workspace_file to remove files.
create_workspaceCreate workspacePOSTCreates an empty workspace (persistent file storage). Pass its id as workspaceId to create_run so runs restore and keep its files, including runs in other sessions.
get_workspaceGet workspaceGETReturns a workspace: its name, whether it is archived, and when it was created and last updated.
rename_workspaceRename workspacePATCHRenames a workspace. An explicit null clears the name; omitting name changes nothing.

Run, 3 operations

create_runCreate runPOSTRuns a Browser Use cloud agent on a natural-language task in a stealth cloud browser and returns the finished run: status, text result, structured output and cost.
get_runGet runGETReturns one run in full: task, status (queued, dispatching, running, completed, failed, cancelled), text result, structured output, error, token counts and LLM cost.
list_runsList runsGETLists the project's agent runs with task, status, result, output, error and cost; filter by sessionId to list the turns of one conversation. Pages with cursor.

Share, 3 operations

enable_session_shareEnable session sharePOSTTurns on a session's read-only share link and returns it. Anyone with shareUrl can see the session's tasks, results, workspace files and recording without an API key.
get_session_shareGet session shareGETReturns a session's read-only share link (active or not), with its URL and view count. Returns nothing if the session was never shared.
update_session_shareUpdate session sharePUTTurns a session's share link on or off. Turning it off stops the link working immediately; turning it on again keeps the same URL.

File, 2 operations

delete_workspace_fileDelete workspace fileDELETEDeletes one file, by its exact path, from a workspace and removes the matching upload record.
list_workspace_filesList workspace filesGETLists the files in a workspace (path, size, last modified), optionally under a path prefix and with download URLs that expire after 60 seconds. Pages with cursor.

Account, 1 operation

get_accountGet accountGETReturns the account behind the API key: credit balance in USD (total, monthly and top-up), concurrent browser limit and active count, plan, free tier flag and project id.

Attachment, 1 operation

list_run_attachmentsList run attachmentsGETLists the upload files attached to a run (name, content type, size, upload status), in the order of the run's attachedFileIds.

Cancel, 1 operation

cancel_runCancel runPOSTCancels an in-flight run and returns it. Idempotent: a run already completed, failed or cancelled is returned unchanged.

Cost, 1 operation

get_session_costGet session costGETReturns everything a session has cost so far, split into LLM, web search, browser time and proxy traffic (USD strings with six decimals), plus proxy MB and run count.

Cost cap, 1 operation

update_session_cost_capUpdate session cost capPATCHChanges the cost cap of the session's active run, in USD, without starting another run.

Download, 1 operation

list_browser_downloadsList browser downloadsGETLists the files a cloud browser downloaded during its session (name, size, date), optionally with presigned download URLs that expire after 15 minutes. Pages with cursor.

Event, 1 operation

list_run_eventsList run eventsGETReads a run's ordered events (lifecycle, model calls, browser readiness, tool activity, artifacts, completion). Pass nextAfter back as after to read only new events.

Purge, 1 operation

purge_sessionPurge sessionPOSTErases a session's run content, workspace files, recordings and downloads (Zero Data Retention projects only; others get 403). A 202 answer means deletion is deferred: call it again after 60 s.

Size, 1 operation

get_workspace_sizeGet workspace sizeGETReturns a workspace's current storage use and its quota, in bytes (the upload allowance is 950 MiB per workspace, 50 MiB per file).

Status, 1 operation

get_run_statusGet run statusGETReturns only a run's status (queued, dispatching, running, completed, failed or cancelled). A cheap read meant for polling; read get_run once the status is final.

Stripe link, 1 operation

get_stripe_link_statusGet stripe link statusGETReturns the project's active Stripe Link wallet connection, if any: whether it is connected, its connectionId, Link email and test or live mode. Connecting is done in the dashboard.

Wallet, 1 operation

list_agentcard_walletsList agentcard walletsGETLists the project's active AgentCard wallets (name, currency, available and held amounts in cents), newest first. Pass a wallet id as agentcardWalletId to create_run to let it pay.

Use Browser Use from Claude Code, Cursor or Codex

Connect Browser Use in LiveContext, create an API key in Settings, then add the LiveContext MCP server to your client. Its Browser Use operations become tools the client can call, next to every other integration you connected.

Claude Code

claude mcp add --transport http livecontext https://livecontext.ai/mcp --header "X-API-Key: YOUR_API_KEY"

Codex (~/.codex/config.toml)

[mcp_servers.livecontext]
url = "https://livecontext.ai/mcp"

[mcp_servers.livecontext.http_headers]
"X-API-Key" = "YOUR_API_KEY"

Cursor, Claude Desktop and other JSON configs

{
  "mcpServers": {
    "livecontext": {
      "type": "http",
      "url": "https://livecontext.ai/mcp",
      "headers": {
        "X-API-Key": "YOUR_API_KEY"
      }
    }
  }
}

Set up the MCP server step by step

How to connect Browser Use

Create your workspace
Sign up free. Nothing to install.
Connect Browser Use once
Add your Browser Use connection in Settings. Every tool from this integration reuses it, and secrets are encrypted at rest and visible only to you.
Describe the job in chat
The workflow builds itself in front of you and picks the endpoints it needs. Drag the steps yourself on the canvas whenever you would rather be explicit.