Test your own app
Point Claude at your dev server or a preview deploy; it reads the page as a snapshot with element refs, clicks, types and takes screenshots.
Integrations / AI agent clients / Claude Code
A browser for Claude Code, one profile per task.
Claude Code can read your code but not your running site. Add the Browz MCP server and it can open your app in a real browser, sign in as a test user and check what changed, each task in its own isolated profile so test sessions never mix with yours.
Terminal
claude mcp add --transport stdio browz -- /Applications/Browz.app/Contents/Resources/browz-mcpAI agent clients
Why use it with Browz.
Sessions that stay apart
Each task opens its own profile: throwaway for a one-off check, or a returning identity that stays signed in to staging as the same test user.
Shared with the team
Add it at project scope and Claude Code writes it to .mcp.json, so everyone who opens the repo gets the same browser tools.
Setup
Set it up.
Install Browz and open it
The MCP server ships inside the app. Download Browz for Mac, Windows or Linux and open it: no account needed for local use.
Add the server for your user
This makes Browz available in every project. Run it in a terminal, not inside Claude Code.
Terminalclaude mcp add --transport stdio browz -- /Applications/Browz.app/Contents/Resources/browz-mcpOr share it with the project
Project scope writes the entry to
.mcp.jsonat the repository root, for everyone who clones it. Use the npm launcher there, since a binary path differs between Mac, Windows and Linux. Claude Code asks each person to approve a project server before it runs.Terminal, in the repositoryclaude mcp add --transport stdio --scope project browz -- npx -y @browz/mcpCheck it is connected
Run
claude mcp listin a terminal, or/mcpinside Claude Code: Browz should show as connected, with its tools.Terminalclaude mcp listRemote, from another computer (beta)
Claude Code can also reach the Browz on your computer through the hosted server. Add it, then run
/mcpto sign in with your Browz account.Terminalclaude mcp add --transport http browz https://mcp.gobrowz.com/mcp
Commands follow Claude Code MCP docs.
Example
Check a sign-up flow as a new user.
Paste a prompt like this into Claude Code with your dev server running. Claude opens a throwaway profile, so the test user never meets your own cookies.
Prompt
Using Browz, open a new throwaway session for "signup check", go to http://localhost:3000/signup, sign up as [email protected] with any valid password, take a screenshot of the page you land on, tell me if anything looks broken, then close the session.Expect calls to browz_session_open, browz_navigate, browz_snapshot, browz_act, browz_screenshot and browz_session_close; the throwaway profile is wiped when the session closes.
Good to know
Notes and tips.
- The snippets use the macOS path of the server inside Browz. On Windows it is
%LOCALAPPDATA%\Programs\Browz\browz-mcp.exe, on Linux/usr/lib/browz/browz-mcp; or usenpx -y @browz/mcp, which finds it. - A profile with a proxy sends its traffic through that proxy; open local dev servers from a profile without one.
- The remote server at
https://mcp.gobrowz.com/mcpis in free public beta: it needs the Browz app open with Remote control on, on a computer with a hardware security chip, and offers the profile, tab and page tools. See what it needs and what it offers. - Browz has to be installed and open on the same computer: the local MCP server drives the app’s own profiles through its local API, and reads the app’s credentials itself, so no token goes in the config.
- A person completes sign-ins, 2FA codes and CAPTCHAs. The agent can pause with
browz_request_humanand hand you the profile’s window, then carry on in the same session. Browz never solves or bypasses CAPTCHAs. - Use agents only where each site’s terms allow automated use, on accounts you are entitled to use. Our acceptable use policy applies to agents the same as to people.
- The MCP server is on every plan, Free included (5 profiles). How many agent sessions can be open at once depends on the plan: see pricing.
FAQ
Claude Code and Browz.
Does Claude Code need my Browz token?
No. Over stdio the server inside Browz reads the app’s own credentials, so the command has no token in it and nothing secret lands in .mcp.json.
What is the difference between user and project scope?
User scope is in your own Claude Code settings and works in every project. Project scope is written to .mcp.json in the repository, so everyone who clones it gets the server after approving it once.
Can Claude Code run several browser tasks at once?
Yes. Each task opens its own session and profile, so parallel tasks never share cookies, storage or a fingerprint. How many sessions can be open at once depends on your plan.
Does it work with Claude Code in a terminal on Linux?
Yes. Browz runs on Linux too; point the command at the server inside the app, /usr/lib/browz/browz-mcp, or use npx -y @browz/mcp.
Get started
Download Browz. Every task, its own browser.
Then, on macOS or Linux, put the MCP server on your PATH
curl -fsSL https://gobrowz.com/agents/install.sh | sh