Docs
Dashboard
SDKs / MCP serverBeta

browser_computer

Run one computer-use action from a model, and see the screen.

Runs ONE action exactly as a computer-use model's tool gives it, in Anthropic's shape ({action: "left_click", coordinate: [x, y]}) or OpenAI's ({type: "click", x, y, button}), on the session's browser page. Then it returns what happened and a screenshot.

Inputs

NameTypeDefaultAllowed valuesDescription
sessionIdrequiredstring––The session's id, from session_create.
actionrequiredobject––The provider's action object, as the model wrote it.
maxWidthinteger––Scale the screenshot down to at most this width (100 to 3840). The action's coordinates are then read in that screenshot's pixels.
screenshotbooleantrue–false skips the screenshot.

This tool works in a session with a browser. On a shell-only session (created with browser: false) it answers browser_disabled. See the three kinds of session.

sessionId is the id that session_create returned. The server keeps no default session. If that session is stopped, this tool resumes it first, as it was (a resume starts a machine again, and billing with it).

Example call

This is what an MCP client sends. You do not write it yourself: ask the model in plain words and it calls the tool.

tools/call
{  "name": "browser_computer",  "arguments": {    "sessionId": "6c1f3c9e-1d2a-4f67-9b0e-1b2c3d4e5f60",    "action": {      "action": "left_click",      "coordinate": [        320,        240      ]    },    "maxWidth": 1024  }}

What it returns

A line saying what happened, the pointer, the scale, the page's title and URL, and the screenshot as an image.

Result
Clicked at (400, 300)Pointer: (320, 240) · scale 0.8 · Example Domain — https://example.com/

When a call fails, the result is marked as an error. Its text starts with the API’s error code, such as unauthorized or session_not_running, followed by its message.

Annotations

The server tells every client what this tool does to the world, so a client can decide what to ask you about before it runs.

AnnotationValue
titleRun a computer-use action
readOnlyHintfalse
destructiveHintfalse
openWorldHinttrue

Good to know

  • It acts on the browser page only, never on the machine's desktop.
  • It is limited to 300 calls a minute per session, and screenshots share that limit with browser_screenshot.

The same thing with the API

EndpointReference
POST /v1/sessions/{id}/browser/computerRun a computer-use action

These docs describe a beta API. Endpoints, fields and SDK methods may change before general availability; breaking changes will be listed in the changelog.