Give Claude Code a real cloud browser
Connect Claude Code to a Nevazu browser over MCP: a real Chrome in the cloud that stays signed in, on its own IP. Nothing to install: you approve the sign-in and watch it work.
Official Claude Code documentation
Claude Code reaches tools through the Model Context Protocol (MCP). Each Nevazu browser has its own hosted MCP server, so Claude Code can take screenshots of the whole screen, click, type, scroll, press shortcuts and upload files in a real Chrome: the same browser, tabs and sign-ins you use in the live view. You watch at the same time and take control back with any click or key.
What it can do
- Read pages the way a person sees them, from screenshots of the real screen: content drawn by scripts, images, charts and layout included.
- Work in the sites you are signed in to, with the session kept between visits.
- See what the browser’s region shows: websites see its dedicated IP and local time zone.
- Upload files from your computer into the browser, for a page’s upload form.
Connect Claude Code
- Create a browser in Nevazu, in the region you want.
- On the browser, choose the robot button (Connect an agent) and, under Use with agents, copy the setup. Paste it into Claude Code: it adds the server itself, as shown below.
- A Nevazu page opens in your web browser. Choose whether it may act or only look, and for how long: 1, 7, 30 or 90 days. Then choose Allow.
Every browser has its own MCP server: its own address (https://connect.nevazu.com/mcp/ followed by that browser’s ID) and its own sign-in, which reaches that browser only. Copy the setup from the browser you want Claude Code to use. For two browsers, add two servers with different names, such as nevazu-work and nevazu-tokyo, and sign in to each.
To add it by hand, the setup for Claude Code looks like this, with your browser’s own address in place of BROWSER_ID:
Run this in a terminal, then restart Claude Code. In Claude Code, /mcp lists the server as needing authentication: choose it, then Authenticate, and the sign-in opens in the user’s web browser. Claude Code keeps the tokens and renews them.
claude mcp add --transport http --scope user nevazu-work https://connect.nevazu.com/mcp/BROWSER_IDThese steps follow Claude Code’s own documentation: Claude Code: Connect Claude Code to tools via MCP ↗
Watching and taking over
While Claude Code is in control, the live view shows a green outline; while you are, a blue one. Any click, key, scroll or paste of yours takes control back before it is carried out, and Claude Code waits until you hand control back. Privacy pause also stops screenshots until you end it.
Questions
Do I need to install anything?
No. The MCP server is hosted by Nevazu and reached over HTTPS; Claude Code only adds its address. There is no key to store: you approve a sign-in instead, and Nevazu emails you each time an agent signs in.
Can Claude Code use the sites I am signed in to?
Yes: it works in the same Chrome and sign-ins you use. That also means it can do anything those accounts allow. Give it only tasks you would do yourself, choose view only where looking is enough, and prefer a separate browser for agent work.
Can Claude Code use more than one browser?
Yes, one MCP server per browser. Each browser has its own server address and its own sign-in, so add one server for each browser you want it to use, under different names. Approving one sign-in never gives access to your other browsers.
What does it cost?
Nevazu is prepaid and billed per minute while the browser runs; its dedicated IP and storage are charged while it exists. A browser an agent is working in stays awake; once the agent stops making requests, it sleeps on its usual timer. Current prices are on the pricing page.
How do I disconnect Claude Code?
In the browser’s Settings → Agent access, or on the Agents page. A disconnected agent loses access at once, even mid-task.