Computer Use

Overview Schema Source Related Servers Discussions Admin

Attach files in browser

browser_upload_file

DestructiveIdempotent

Attach local files to a browser tab's file input using a snapshot index, without opening a file picker or submitting the form.

Instructions

Attach local files to a file input in one of the agent's tabs, by its browser_snapshot index, as if the user had picked them: Chrome sets the input's files and the page gets its usual input and change events. No file picker opens, the tab is not selected or raised, and no keyboard or mouse input is used. Snapshots list file inputs as input[file], including ones a page hides behind a styled upload button; the index of a label for one works too, and any other element is refused. Paths are absolute paths of regular files on the computer running Chrome and this server, which may not be the machine you run on; each is checked before anything is sent. Several paths need an input that takes several files (multiple). This only stages the files in the input: it never submits the form, and the site has not received them until the page shows it has. Submit or press the page's upload button yourself when the user wants that, then check the page. The reply names the files the input now holds. Never click a file input or its upload button to open the picker: the native dialog cannot be driven from here.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "tab_id": "<tab_id>",
      "index": "<index>",
      "paths": "<paths>"
    },
    "name": "browser_upload_file"
  }
}

Input Schema

Name	Required	Description
`index`	Yes	Element index of the file input (input[file]), or of a label for one, from the latest browser_snapshot of this tab.
`paths`	Yes	Absolute paths of the files to attach, on the computer running Chrome and this server. Each must exist and be a regular file, not a folder. More than one needs an input with the multiple attribute. Replaces whatever files the input held.
`tab_id`	Yes	tab_id of one of the agent's tabs, from browser_open_tab or browser_list_tabs
`session_id`	No	Stable task or thread ID. Pass the same value on every browser call for this task. Different IDs isolate tabs and cleanup within one MCP process. Omit for the default process session.
`return_state`	No	Append a current snapshot and readiness status after acting. Without wait_for_selector, only a short pause and bounded navigation wait are used; SPA readiness is not guaranteed. Snapshot indices replace earlier ones. A timeout does not mean the action failed: re-observe with browser_snapshot before repeating it.
`wait_timeout_ms`	No	Bound for wait_for_selector in milliseconds (default 8000, maximum 10000; 0 checks once). Requires wait_for_selector. Timeout preserves action success and returns current state. Use browser_snapshot with the same wait to re-observe without repeating an action.
`wait_for_selector`	No	Wait for any visible element matching this CSS selector in the page, including open shadow roots (web components); child frames are not searched, and a timeout says how many were visible. On actions requires return_state=true. Returns readiness met, timeout or error; only this condition is checked, not general app readiness. Prefer a selector specific to the expected new UI.

Schema Changelog

Changes observed during successful MCP inspections.

Addedv0.13.0Nov 7, 2026

TDQS

A4.8/5.0

Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Beyond the annotations it discloses a lot: no picker opens, the tab is not selected or raised, no keyboard or mouse input is used, paths are validated before anything is sent, and the action only stages files without submitting. It also explains the reply contents and the label/hidden-input acceptance rules, all consistent with destructiveHint=true (the input's contents are replaced) and idempotentHint=true.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Purpose is front-loaded and every sentence carries operational information, but the prose is dense and runs long for a single tool, with mild overlap between 'as if the user had picked them' and the following sentence describing Chrome setting files and firing input/change events.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a 7-parameter, non-read-only tool with no output schema, the definition covers the full lifecycle: preconditions on paths and the input element, what the tool does and does not do, what the response reports, and the recovery step of submitting and re-checking the page. Nothing an agent needs to invoke it correctly is missing.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100%, so the schema already documents each parameter. The description still adds meaning: index may point at a label rather than the input while any other element is refused, multiple paths require an input with the multiple attribute, and paths live on the Chrome host machine rather than the agent's machine — a non-obvious constraint worth surfacing.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The opening sentence states a specific verb and resource ('Attach local files to a file input'), plus the addressing mechanism (the agent's tab, by browser_snapshot index). It is immediately distinguishable from siblings like browser_fill, browser_click, or browser_type, which an agent might otherwise guess at for uploads.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

It says explicitly when this applies (staging files into an input) and what to do afterwards ('Submit or press the page's upload button yourself... then check the page'), and gives a hard negative rule: never click a file input to open the picker because the native dialog cannot be driven. The multiple-files precondition is also stated as a usage constraint.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.