AI agents / Browz for OpenAI agents

A browser for your OpenAI agents, one profile per task.

You decide exactly what each agent may do: its own key, its own profiles, only the sites you allow, sign-ins it can use on your computer but never read, and a person for the login. Agents you build with the OpenAI Agents SDK can browse in Browz through MCP. Each run gets an isolated profile with its own fingerprint, proxy and cookies, sessions carry over to the next run, and the same tools work from the Responses API through the remote server, in beta. Underneath, every profile’s fingerprint is set in the browser engine itself, not by a script on top, and you can test it yourself (engine-level vs injected scripts).

Set it up Download Browz

Agents SDK (Python), running on the computer with Browz

import asyncio
from agents import Agent, Runner
from agents.mcp import MCPServerStdio

async def main():
    async with MCPServerStdio(name="browz", params={"command": "/Applications/Browz.app/Contents/Resources/browz-mcp", "args": []}) as browz:
        agent = Agent(name="Browz agent", mcp_servers=[browz])
        result = await Runner.run(agent, "Using Browz: create a new profile called agent-check, start it, open https://example.com, tell me the page title, take a screenshot, then stop the session and delete the profile.")
        print(result.final_output)

asyncio.run(main())

Browz for OpenAI agents

What your agent gets.

Agents SDK, on your computer

MCPServerStdio starts the server inside Browz, and the agent gets every Browz tool. Nothing to paste: it reads the app’s own credentials.

OpenAI setup 02

Responses API, from the cloud

OpenAI can call the Browz on your computer through the remote server at https://mcp.gobrowz.com/mcp, with the same tool names, so a prompt works either way. In beta: tell us how it works for you from Settings › Help and support.

Beta

Remote setup

Sessions per run

Open a throwaway profile for a one-off task, or a returning one for an identity the agent works as every day, with its sign-in kept.

Measured, then published

In our published benchmark the same AI agent finished 46 of 50 public web tasks through Browz and 38 of 50 through a plain browser MCP server; on sites that challenge automated browsers it was 14 of 18 against 9 of 18.

See the benchmark

Setup

Working in minutes.

These are the same snippets as the setup docs and the Connect an AI agent screen in the app.

Install Browz and open it
Download Browz on the computer that runs your agent.
Install the Agents SDK
Install the OpenAI Agents SDK for Python, set OPENAI_API_KEY, and save the script below.
Run it
The agent should answer “Example Domain” and leave no profile behind.

Install Browz, open it, and your agent can start: no account needed for local use. Keep the app open while your agent works; it can stay on its sign-in screen.

Agents SDK (Python) Agents SDK (Python), running on the computer with Browz

Agents SDK (Python)

import asyncio
from agents import Agent, Runner
from agents.mcp import MCPServerStdio

async def main():
    async with MCPServerStdio(name="browz", params={"command": "/Applications/Browz.app/Contents/Resources/browz-mcp", "args": []}) as browz:
        agent = Agent(name="Browz agent", mcp_servers=[browz])
        result = await Runner.run(agent, "Using Browz: create a new profile called agent-check, start it, open https://example.com, tell me the page title, take a screenshot, then stop the session and delete the profile.")
        print(result.final_output)

asyncio.run(main())

The snippets use the macOS path to the server inside Browz, /Applications/Browz.app/Contents/Resources/browz-mcp; on Windows and Linux the paths are in the setup docs. Give each agent its own key, limited to the profiles and sites it needs, and let it sign in without seeing the password. The remote MCP server for hosted agents is in Beta: free for every account, Free included, within the beta allowance, and it needs the Browz app open with Remote control on, on a computer with a hardware security chip. See what it needs and its allowance. Use agents only where each site’s terms allow it: our acceptable use policy applies to agents the same as to people.

FAQ

About Browz for OpenAI agents.

Can I use the Responses API instead?

Yes, through the remote MCP server, which is in free public beta. The setup docs show the Responses API and the hosted MCP tool of the Agents SDK, each with your Browz credentials read from an environment variable.

Which models work?

Any model the Agents SDK runs. Browz is the browser; the model is your choice.

Can several agents run at once?

Yes. Each session gets its own profile, so parallel agents never share cookies, storage or a fingerprint.

More for agent builders

Your agent. Its own browser.

Download Browz See pricing

For Mac, Windows and Linux. Free for 5 profiles.

A browser for your OpenAI agents, one profile per task.

What your agent gets.

Agents SDK, on your computer

Responses API, from the cloud

Sessions per run

Measured, then published

Working in minutes.

Install Browz and open it

Install the Agents SDK

Run it

About Browz for OpenAI agents.

Other guides.

Browz for AI agents

Anti-detect browser for AI agents

MCP browser

Browz for Claude

Your agent. Its own browser.