RELATED 01
Browser Use Cloud MCP
OFFICIAL SERVERHosted endpoint that hands whole web tasks to Browser Use's cloud browser agent and returns the result
40popularity score
Page Agent MCP
Lets an assistant hand browser tasks in plain language to Page Agent running in your own Chrome.
By Alibaba JavaScript Local · stdio Browser automation
46 popularity score ★ 30k ↓ 167/wk
Connect your client
Local (npm)
Add in one click
Opens the app, which asks before it adds the server. Then replace each <NAME> in its settings with your value.
Or add it by command or config
It needs LLM_BASE_URL, LLM_API_KEY, LLM_MODEL_NAME. Replace each <NAME> with your value.
Claude CodeTerminal
claude mcp add page-agent -e LLM_BASE_URL=<LLM_BASE_URL> -e LLM_API_KEY=<LLM_API_KEY> -e LLM_MODEL_NAME=<LLM_MODEL_NAME> -- npx -y @page-agent/mcpClaude Desktopclaude_desktop_config.json
{
"mcpServers": {
"page-agent": {
"command": "npx",
"args": [
"-y",
"@page-agent/mcp"
],
"env": {
"LLM_BASE_URL": "<LLM_BASE_URL>",
"LLM_API_KEY": "<LLM_API_KEY>",
"LLM_MODEL_NAME": "<LLM_MODEL_NAME>"
}
}
}
}Cursor~/.cursor/mcp.json
{
"mcpServers": {
"page-agent": {
"command": "npx",
"args": [
"-y",
"@page-agent/mcp"
],
"env": {
"LLM_BASE_URL": "<LLM_BASE_URL>",
"LLM_API_KEY": "<LLM_API_KEY>",
"LLM_MODEL_NAME": "<LLM_MODEL_NAME>"
}
}
}
}VS Code.vscode/mcp.json
{
"servers": {
"page-agent": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@page-agent/mcp"
],
"env": {
"LLM_BASE_URL": "<LLM_BASE_URL>",
"LLM_API_KEY": "<LLM_API_KEY>",
"LLM_MODEL_NAME": "<LLM_MODEL_NAME>"
}
}
}
}Codex~/.codex/config.toml
[mcp_servers.page-agent]
command = "npx"
args = ["-y", "@page-agent/mcp"]
[mcp_servers.page-agent.env]
LLM_BASE_URL = "<LLM_BASE_URL>"
LLM_API_KEY = "<LLM_API_KEY>"
LLM_MODEL_NAME = "<LLM_MODEL_NAME>"Gemini CLITerminal
gemini mcp add --scope user -e LLM_BASE_URL=<LLM_BASE_URL> -e LLM_API_KEY=<LLM_API_KEY> -e LLM_MODEL_NAME=<LLM_MODEL_NAME> page-agent npx -- -y @page-agent/mcpSettings 4
LLM_BASE_URL |
Required: Base URL of the OpenAI-compatible LLM API Page Agent uses. |
|---|---|
LLM_API_KEY |
Required: API key for that LLM API. |
LLM_MODEL_NAME |
Required: Model name to use. |
PORT |
Optional: Port for the local HTTP and WebSocket bridge (default 38401). |
What it does
Page Agent, from Alibaba, is a JavaScript agent that runs inside a web page and controls the interface from natural-language instructions. Its MCP server (@page-agent/mcp, marked beta) lets MCP clients such as Claude Desktop, Cursor or Copilot use it to drive your browser: the assistant sends a task described in words, Page Agent carries it out in Chrome, and the result comes back.
The server has three tools: run a browser task (it blocks until the task finishes), check whether the browser is connected and busy, and stop the current task. It does not expose clicks and typing as separate tools; Page Agent's own model decides the steps.
It needs Node.js 20 or newer, the Page Agent extension installed in Chrome, and an OpenAI-compatible LLM API (LLM_BASE_URL, LLM_API_KEY, LLM_MODEL_NAME), which Page Agent uses to plan its actions. Started with npx -y @page-agent/mcp, it opens a local HTTP and WebSocket bridge (default port 38401) and a launcher page that connects the extension.
What it needs
| Runtime | Node.js |
|---|
Tools 3
execute_task- Execute a browser task described in natural language; blocks until it is done.
get_status- Return whether the browser hub is connected and busy.
stop_task- Stop the currently running task.