BrowserCity MCP

Give your agent a real browser over MCP

Connect your favorite MCP client to BrowserCity and let your agent open pages, click through workflows, inspect DOMs, and return clean markdown. No local browser infrastructure required.

Copy a quickstart Browse integrations

What it enables

Give your agent hands and eyes on the web

MCP is the control surface. We provide the live, hosted browser session behind it. Now your agent can actively browse, click, and extract instead of just reading static text.

Open a managed browser session

Start Chromium from your agent with BrowserCity handling browser hosting, lifecycle, and connection details.

Navigate and interact through tools

Let the agent call tools for navigation, clicking, typing, form filling, tab control, screenshots, and page inspection.

Return structured page context

Capture markdown, HTML, snapshots, and selected page state so the next agent step has usable web context.

How it works

One remote server, full browser control

Your MCP client sends authenticated tool calls to BrowserCity. We instantly spin up or reuse a cloud browser session, execute the requested action, and return the result.

MCP client reads your config and connects to https://mcp.browser.city/mcp.

The Authorization header maps the request to your BrowserCity organization and token scope.

The agent calls tools such as browser_open, browser_navigate, and browser_markdown.

BrowserCity runs the browser work remotely and returns snapshots, content, files, or action results.

Quickstarts

Copy-paste setup for supported MCP clients

These examples use the current BrowserCity endpoint and client-specific remote MCP syntax from official docs. If your client is not listed, use the generic connection details above.

CLI install

Claude Code

Full guide

Add BrowserCity as a remote HTTP MCP server and pass your API token as a bearer header.

Terminal

claude mcp add --transport http browsercity https://mcp.browser.city/mcp \n  --header "Authorization: Bearer $BROWSERCITY_API_KEY"

Official Claude Code MCP docs

Project or global config

Cursor

Full guide

Use Cursor's remote MCP config with an environment-expanded Authorization header.

.cursor/mcp.json

{  "mcpServers": {    "browsercity": {      "url": "https://mcp.browser.city/mcp",      "headers": { "Authorization": "Bearer ${env:BROWSERCITY_API_KEY}" }    }  }}

Official Cursor MCP docs

Cascade config

Windsurf

Full guide

Point Cascade at the remote Streamable HTTP endpoint from Windsurf's MCP config.

~/.codeium/windsurf/mcp_config.json

{  "mcpServers": {    "browsercity": {      "serverUrl": "https://mcp.browser.city/mcp",      "headers": { "Authorization": "Bearer ${env:BROWSERCITY_API_KEY}" }    }  }}

Official Windsurf MCP docs

Available tools

Start with the common browser loop

Tool availability can vary by client policy, but BrowserCity exposes a browser toolset covering session lifecycle, navigation, page content, interaction, and capture.

browser_open / browser_closebrowser_navigate / browser_navigate_backbrowser_click / browser_type / browser_fill_formbrowser_snapshot / browser_markdown / browser_htmlbrowser_tabs / browser_resize / browser_take_screenshot

Use MCP when

Your agent needs to decide the next browser action

MCP fits multi-step workflows where the agent needs feedback from the page before choosing the next click, form entry, navigation, or extraction step.

Use REST when

You already know the browser task

For deterministic URL-to-markdown extraction or your own automation code, use the Request API or Sessions API directly.

Read the API docs

[ 06 / 06 ] — Get Started

Give your AI agents the web.

Start for free. No credit card required. Private sessions by default.

Start Free Read Docs Join Discord