Browser User

Enterprise

Browser Use

Product overview

Name of Agent: Browser User
Short description of agent: "AI browser agent. Repetitive work is dead. Browser Use empowers anyone to automate repetitive online tasks, no code required. No barriers. Simply tell it what you want done." (link, archived)
Date of release: 01/2025 (link, archived) for cloud interface
Advertised use: "extracts data from any website", "accurately fills out online forms using data from any source" (source, archived)
Monetisation/Usage price: 250, 500, more compute or pay as you go with credits
Who is using it?: enterprise companies, can be used through cloud, api, self hosted, or as MCP server
Category: Enterprise

Company & accountability

Developer: Browser Use
Name of legal entity: Browser Use Inc.
Place of legal incorporation: Delaware (link, archived). SF is physical location (link, archived) of office.
For profit company?: Yes
Parent company?: Not applicable
Governance documents analysis: Customer TOS (link, archived)
AI safety/trust framework: None found
Compliance with existing standards: SOC 2 (link, archived)

Technical capabilities & system architecture

Model specifications: Supports a range of frontier models (cite, archived). Reasoning models available.
Observation space: Textual representation of websites (DOM) (link, archived), Internet access (through DOM representation), and returns from executed Python code (link, archived)
Action space: Actions in browser and text back to user (link, archived)
Memory architecture: Tool returns can be set to always be put in model context (link, archived). Docs state the agents "maintains memory" across tasks (link, archived).
User interface and interaction design: [non-cloud] Terminal and code files. Specify what to do in the file and inspect agent outputs in the terminal [cloud] chatbot
User roles: [non-cloud] Designer (user architects and codes the agent); Operator + Executor + Examiner (user can test out the agent and iterate on it based on agent's performance) [cloud]: Operator + Executor
Component accessibility: Open source framework (link)

Autonomy & control

Autonomy level and planning depth: L5: does all planning and execution, have not seen affordances for any kind of user interaction before/during execution besides assigning the task
User approval requirements for different decision types: None
Execution monitoring, traces, and transparency: All actions are
Emergency stop and shut down mechanisms and user control: User can pause/stop the agent at any time
Usage monitoring and statistics and patterns: Stats and usage are available in the Analytics and Agent Sessions tabs of the cloud console

Ecosystem interaction

Identify to humans?: None, Because Browser Use traffic looks like a normal interactive browser session (clicks, form fills, JS rendering) that sites would see as human‑like, there is no disclosure about an AI assistant automating browsing work.
Identifies technically?: None, "Our browser infrastructure is stealth by default," suggests to me that Browser Use traffic does not try to make itself identifiable and looks like traffic coming from a user's browser (link, archived) Browser Use’s official marketing emphasizes “Browse the web like a human” and does not mention any built-in mechanism to disclose AI automation to third-party sites/non-user humans.
Interoperability standards and integrations: None, but provides an MCP server for other other AI assistant to use their browser automation tools (link, archived)
Web conduct: - "Bypass any captcha or anti-bot systems automatically," suggests to me that they don't obey robots.txt ((link, archived), (link, archived)) - "Browses the web like a human" (link, archived)

Safety, evaluation & impact

Technical guardrails and safety measures: None found
Sandboxing and containment approaches: Sandbox appears to be recommended but not required (see developer docs (link, archived))
What types of risks were evaluated?: None found
(Internal) safety evaluations and results: None found
Third-party testing, audits, and red-teaming: None found
Benchmark performance and demonstrated capabilities: None found
Bug bounty programmes and vulnerability disclosure: Yes (link), Browser Use provides a vulnerability disclosure/reporting process via GitHub Security Advisories
Any known incidents?: None found
← Back to 2025 Index