dev-browser-mcp
Enables AI agents to control a persistent Chromium browser or attach to an existing Chrome with sandboxed JavaScript and structured tools for web interaction and automation.
README
dev-browser-mcp
MCP server for SawyerHood/dev-browser. Agents drive a persistent Chromium (or an already-running Chrome) with sandboxed JavaScript and a small set of structured tools.
Repository: github.com/kamkozlowski/dev-browser-mcp
Scripts run in QuickJS, not Node.js. Named pages persist across tool calls through the dev-browser daemon.
dev-browser (CLI + Playwright Chromium) is a dependency of this package. npm install downloads the CLI and installs Chromium. No global dev-browser install is required. Set DEV_BROWSER_SKIP_CHROMIUM=1 to skip the Chromium download (for example when you only attach to an existing Chrome). DEV_BROWSER_BIN still overrides the bundled CLI.
Install
Clone and build:
git clone https://github.com/kamkozlowski/dev-browser-mcp.git
cd dev-browser-mcp
npm install
npm run build
Or install from the GitHub tarball:
npm pack github:kamkozlowski/dev-browser-mcp
tar -xzf dev-browser-mcp-*.tgz
cd package
npm install
npm run build
Cursor config
Add to ~/.cursor/mcp.json (or project .cursor/mcp.json).
Recommended: bootstrap from GitHub (cache under ~/.cache)
No local clone needed. On first launch Cursor downloads the package, installs dependencies, builds, and starts the server:
{
"mcpServers": {
"dev-browser": {
"command": "bash",
"args": [
"/ABS/PATH/TO/dev-browser-mcp/scripts/bootstrap-mcp.sh"
],
"env": {
"DEV_BROWSER_HEADLESS": "true"
}
}
}
}
If you do not keep a local clone, use this one-liner (same logic as scripts/bootstrap-mcp.sh):
{
"mcpServers": {
"dev-browser": {
"command": "bash",
"args": [
"-lc",
"CACHE=\"${XDG_CACHE_HOME:-$HOME/.cache}/dev-browser-mcp\"; LOCK=\"$CACHE/.install.lock\"; PKG=\"$CACHE/package\"; ENTRY=\"$PKG/dist/index.js\"; mkdir -p \"$CACHE\"; install() { cd \"$CACHE\" && rm -rf package dev-browser-mcp-*.tgz && npm pack github:kamkozlowski/dev-browser-mcp >/dev/null && tar -xzf dev-browser-mcp-*.tgz && rm -f dev-browser-mcp-*.tgz && cd package && npm install && npm run build; }; if [ ! -f \"$ENTRY\" ]; then ( flock -n 9 || flock 9; [ -f \"$ENTRY\" ] || install ) 9>\"$LOCK\"; fi; [ -f \"$ENTRY\" ] || { echo \"Install failed; rm -rf $CACHE\" >&2; exit 1; }; cd \"$PKG\" && exec node dist/index.js"
],
"env": {
"DEV_BROWSER_HEADLESS": "true"
}
}
}
}
If the server fails to start with Cannot find module .../dist/index.js, remove the broken cache and retry:
rm -rf "${XDG_CACHE_HOME:-$HOME/.cache}/dev-browser-mcp"
To attach to your running Chrome instead of launching Chromium:
{
"mcpServers": {
"dev-browser": {
"command": "bash",
"args": [
"-lc",
"CACHE=\"${XDG_CACHE_HOME:-$HOME/.cache}/dev-browser-mcp\"; LOCK=\"$CACHE/.install.lock\"; PKG=\"$CACHE/package\"; ENTRY=\"$PKG/dist/index.js\"; mkdir -p \"$CACHE\"; install() { cd \"$CACHE\" && rm -rf package dev-browser-mcp-*.tgz && npm pack github:kamkozlowski/dev-browser-mcp >/dev/null && tar -xzf dev-browser-mcp-*.tgz && rm -f dev-browser-mcp-*.tgz && cd package && npm install && npm run build; }; if [ ! -f \"$ENTRY\" ]; then ( flock -n 9 || flock 9; [ -f \"$ENTRY\" ] || install ) 9>\"$LOCK\"; fi; [ -f \"$ENTRY\" ] || { echo \"Install failed; rm -rf $CACHE\" >&2; exit 1; }; cd \"$PKG\" && exec node dist/index.js"
],
"env": {
"DEV_BROWSER_CONNECT": "auto",
"DEV_BROWSER_SKIP_CHROMIUM": "true"
}
}
}
}
Enable remote debugging in Chrome at chrome://inspect/#remote-debugging, or launch with --remote-debugging-port=9222.
Local clone (simple)
After npm install and npm run build in your checkout:
{
"mcpServers": {
"dev-browser": {
"command": "node",
"args": ["/ABS/PATH/TO/dev-browser-mcp/dist/index.js"],
"env": {
"DEV_BROWSER_HEADLESS": "true"
}
}
}
}
Local clone (self-bootstrapping, bash)
Cursor installs and builds on first launch if dist/ is missing:
{
"mcpServers": {
"dev-browser": {
"command": "bash",
"args": [
"-lc",
"REPO=\"/ABS/PATH/TO/dev-browser-mcp\"; if [ ! -f \"$REPO/dist/index.js\" ]; then cd \"$REPO\" && npm install && npm run build; fi; cd \"$REPO\" && exec node dist/index.js"
],
"env": {
"DEV_BROWSER_HEADLESS": "true"
}
}
}
}
Replace REPO with your checkout path.
Development (TypeScript, no build)
{
"mcpServers": {
"dev-browser": {
"command": "bash",
"args": [
"-lc",
"REPO=\"/ABS/PATH/TO/dev-browser-mcp\"; cd \"$REPO\" && exec npx --yes tsx src/index.ts"
],
"env": {
"DEV_BROWSER_HEADLESS": "true"
}
}
}
}
Replace REPO with your checkout path.
Tools
| Tool | Purpose |
|---|---|
dev_browser_run |
Main tool. Execute a sandboxed Playwright script. |
dev_browser_list_pages |
List tabs: {id, url, title, name} |
dev_browser_page_open |
Get/create a named page (or attach by targetId) |
dev_browser_close_page |
Close a named page |
dev_browser_goto |
Navigate |
dev_browser_snapshot |
AI accessibility snapshot (snapshotForAI) |
dev_browser_screenshot |
PNG screenshot (also returned as an image) |
dev_browser_click |
Click by CSS selector or snapshot ref |
dev_browser_fill |
Fill an input |
dev_browser_type |
Type character by character |
dev_browser_evaluate |
eval JavaScript in the page |
dev_browser_wait_for_selector |
Wait for a selector |
dev_browser_status |
Daemon status |
dev_browser_browsers |
Managed browser instances |
dev_browser_stop |
Stop the daemon |
dev_browser_install |
Install Playwright Chromium |
Read dev-browser://guide for the full sandbox API, including page.cua (vision) and page.domCua (DOM ids).
Environment
| Variable | Meaning |
|---|---|
DEV_BROWSER_BIN |
Override the bundled dev-browser CLI path |
DEV_BROWSER_SKIP_CHROMIUM |
true to skip postinstall / auto Chromium setup |
DEV_BROWSER_BROWSER |
Named daemon browser instance (default: default) |
DEV_BROWSER_HEADLESS |
true to launch Chromium without a window |
DEV_BROWSER_CONNECT |
auto / true to attach to Chrome, or a CDP URL |
DEV_BROWSER_IGNORE_HTTPS_ERRORS |
true for self-signed certs |
DEV_BROWSER_TIMEOUT |
Script timeout in seconds (default: 30) |
DEV_BROWSER_IDLE_TIMEOUT |
Close idle launched browsers, e.g. 5m |
Per-call browser, connect, headless, ignoreHttpsErrors, and timeoutSeconds override the environment.
How it differs from Playwright MCP
dev-browser is built around one sandboxed script per decision, not a long chain of atomic MCP calls. Use dev_browser_run for anything beyond a single goto/click/fill. Convenience tools exist so common inspect/act steps stay structured.
Unlike benkraus/dev-browser-mcp (Chrome extension + CDP relay), this server wraps the official dev-browser CLI with its QuickJS sandbox and persistent daemon pages.
Development
npm test
npm run typecheck
npm run build
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.