Edge DevTools MCP
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Edge DevTools MCPnavigate to example.com and take a screenshot"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Edge DevTools MCP
Technical Overview & Metadata
Package Name:
edge-devtools-mcpSpecification: Model Context Protocol (MCP) Standard Server
Target Engine: Microsoft Edge (Chromium runtime) & Chrome DevTools Protocol (CDP)
Primary Runtime: Node.js >= 18.0.0
Supported Platforms: Windows 10/11, macOS (Apple Silicon / Intel), Linux (x64 / arm64)
Supported AI Clients: Google Antigravity, Claude Code (CLI & Desktop), Cursor, Visual Studio Code (Roo Code, Cline, Continue), Windsurf, Codex
About for Search Engines & Generative AI (SEO / GEO):
edge-devtools-mcpis an open-source, production-grade Model Context Protocol server that bridges Large Language Model (LLM) agents with Microsoft Edge. It provides full-spectrum browser control, ranging from lightweight accessibility tree navigation (saving up to 90% in token consumption) to raw Chrome DevTools Protocol (CDP) invocation for network interception, performance profiling, DOM mutation, cookie manipulation, and console telemetry.
Related MCP server: Fast Browser MCP
Architectural Design
The server exposes a four-tiered tool hierarchy designed for optimal agent reasoning and token efficiency:
flowchart TD
Client["AI Agent / MCP Client<br/>(Antigravity / Claude Code / Cursor / VS Code)"]
subgraph EdgeMCP["Edge DevTools MCP Server (stdio)"]
Router["MCP Request Router & Validator (Zod)"]
subgraph Layers["Execution Capabilities"]
L1["Layer 1: Session & Navigation<br/>(launch, connect, navigate, tabs, close)"]
L2["Layer 2: User Interactions<br/>(click, fill, press_key, hover, scroll)"]
L3["Layer 3: Token-Optimized Inspection<br/>(accessibility tree, screenshot, eval, html)"]
L4["Layer 4: DevTools Protocol (CDP)<br/>(cdp_execute, network, console, cookies, emulation)"]
end
Hybrid["Hybrid Connection Engine<br/>(Auto-detects active Edge port 9222 or launches fresh)"]
end
EdgeInstance["Microsoft Edge Browser<br/>(Local binary / Persistent profile / CDP Session)"]
Client <--> Router
Router --> L1 & L2 & L3 & L4
L1 & L2 & L3 & L4 --> Hybrid
Hybrid <--> EdgeInstanceKey Capabilities
Hybrid Connection Lifecycle: Automatically attaches to an existing Microsoft Edge instance running with
--remote-debugging-port=9222to leverage already authenticated sessions, cookies, and enterprise profiles. If no instance is active, it launches an isolated browser session automatically.Token-Efficient Accessibility Tree: Implements semantic aria snapshot extraction (
edge_get_accessibility_snapshot), providing the LLM with structured element hierarchies and accessibility roles instead of polluting the context window with raw HTML.100% Raw CDP Power: Provides
edge_cdp_executeto execute any native DevTools domain method (Network.*,Page.*,DOM.*,Storage.*,Emulation.*,Performance.*) with custom payloads.Real-Time Telemetry Capture: Captures console output (errors, warnings, logs) and network activity (HTTP status, method, timing, headers) without requiring external proxy setups.
Installation & Setup
You can run the server directly via npx without global installation:
npx -y edge-devtools-mcpOr install globally:
npm install -g edge-devtools-mcpClient Configuration Guides
1. Google Antigravity
Add the server configuration to your MCP configuration file (.gemini/antigravity/mcp.json or through settings):
{
"mcpServers": {
"edge-devtools": {
"command": "npx",
"args": ["-y", "edge-devtools-mcp"]
}
}
}2. Claude Code CLI
Install directly using the Claude Code MCP command:
claude mcp add edge-devtools -- npx -y edge-devtools-mcp3. Claude Desktop
Add to claude_desktop_config.json:
Windows:
%APPDATA%\Claude\claude_desktop_config.jsonmacOS:
~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"edge-devtools": {
"command": "npx",
"args": ["-y", "edge-devtools-mcp"]
}
}
}4. Visual Studio Code (Roo Code / Cline / Continue)
Add to your global or workspace mcpSettings.json or extension configuration:
{
"mcpServers": {
"edge-devtools": {
"command": "npx",
"args": ["-y", "edge-devtools-mcp"]
}
}
}5. Cursor
Configure in Settings > Features > MCP or create .cursor/mcp.json:
{
"mcpServers": {
"edge-devtools": {
"command": "npx",
"args": ["-y", "edge-devtools-mcp"]
}
}
}6. Windsurf / Codex
Add to your MCP configuration file:
{
"mcpServers": {
"edge-devtools": {
"command": "npx",
"args": ["-y", "edge-devtools-mcp"]
}
}
}Tools Reference
Navigation & Session Management
Tool Name | Parameters | Description |
|
| Launches Microsoft Edge with optional persistent profile, custom port, or headless mode. |
|
| Connects to an existing Edge debugging session over CDP (default: |
|
| Navigates the active tab to a URL with configurable wait conditions. |
|
| Lists tabs, creates a new tab, switches active tab, or closes an open tab. |
| none | Closes all open pages and terminates the Microsoft Edge process. |
User Interaction
Tool Name | Parameters | Description |
|
| Clicks an element by CSS selector, text match, or XPath. |
|
| Fills input or textarea element with text and optionally presses Enter. |
|
| Sends keyboard keystrokes ( |
|
| Hovers the cursor over a target element to trigger tooltips or menus. |
|
| Scrolls by pixel offsets or scrolls a specific element into view. |
|
| Selects options in dropdown |
Inspection & Token-Optimized Extraction
Tool Name | Parameters | Description |
|
| Recommended for Agents. Returns semantic accessibility tree (roles, names, states) to conserve tokens. |
|
| Returns raw HTML of page or element with configurable character limit. |
|
| Captures screenshot as base64 image or saves directly to a file path. |
|
| Evaluates arbitrary JavaScript inside the page and returns serialized result. |
Edge DevTools & CDP Low-Level Control
Tool Name | Parameters | Description |
|
| Executes any native Chrome/Edge DevTools Protocol command (e.g. |
|
| Inspects captured browser console messages and uncaught runtime errors. |
|
| Inspects network requests, HTTP status codes, and request methods. |
|
| Inspects, injects, or purges session cookies across the browser context. |
|
| Overrides viewport size, mobile touch emulation, and light/dark color scheme. |
Advanced Usage Scenarios
Connecting to Your Daily Authenticated Edge Session
To let the agent use your existing logins (e.g., GitHub, internal dashboards, portals):
Close existing Edge instances or start Edge with a dedicated debug port:
# Windows PowerShell Start-Process msedge.exe -ArgumentList "--remote-debugging-port=9222" # macOS Terminal /Applications/Microsoft\ Edge.app/Contents/MacOS/Microsoft\ Edge --remote-debugging-port=9222 # Linux microsoft-edge --remote-debugging-port=9222The agent automatically detects and attaches to
http://localhost:9222.
Example: Running Raw CDP Commands via Agent
The agent can invoke advanced profiling and network interception using edge_cdp_execute:
{
"method": "Network.emulateNetworkConditions",
"params": {
"offline": false,
"downloadThroughput": 500000,
"uploadThroughput": 500000,
"latency": 50
}
}Local Development & Testing
# Clone the repository
git clone https://github.com/dannymaaz/edge-devtools-mcp.git
cd edge-devtools-mcp
# Install dependencies
npm install
# Run automated tests
npm test
# Build distribution bundle
npm run build
# Test using the official MCP Inspector
npx @modelcontextprotocol/inspector node dist/bin/cli.jsVerification & Quality Assurance
Rigorous Schemas: All tool inputs and outputs are validated with Zod and TypeScript.
Cross-Platform Path Resolution: Automatically locates Microsoft Edge on Windows (
Program Files,Local AppData), macOS (/Applications), and Linux distributions.Resource Cleanup: Guaranteed browser and socket cleanup on
SIGINTandSIGTERMsignals.
License
This project is licensed under the MIT License.
This server cannot be deployed
Maintenance
Related MCP Connectors
Control a real browser from AI agents: navigate, click, type, screenshot, read the page.
Provides cloud browser automation capabilities using Stagehand and Browserbase, enabling LLMs to i…
Undetectable cloud browser sessions for AI agents and scrapers. Navigate, extract, click, captcha.
Hosted browser for agents: open pages, search the web, extract data, read PDFs and take screenshots.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceEnables control of Microsoft Edge browser via the DevTools protocol, supporting page navigation, screenshots, DOM manipulation, JavaScript execution, and network monitoring.28 npm14MIT
- AlicenseBqualityAmaintenanceEnables AI agents and developers to automate real Chrome/Chromium browsers through the Chrome DevTools Protocol, with token-efficient snapshots, multi-action batching, shadow DOM traversal, system-page control, extension management, and network/WebSocket inspection.691MIT
- AlicenseNot gradedqualityCmaintenanceEnables MCP clients to drive a running Microsoft Edge browser over the Chrome DevTools Protocol, exposing 16 tools for navigation, clicking, typing, screenshots, page-content reading, and tab management. Runs in stdio mode for per-client processes or as a shared HTTP endpoint so multiple agents can operate the same browser session.MIT
- FlicenseNot gradedqualityNot gradedmaintenanceEnables AI agents to browse and interact with web pages through a persistent headless Chromium session, serving compact structured orientations, targeted element queries, and full detail only when explicitly requested. It supports navigation, observation, form filling, monitoring, and session management while cutting token overhead by up to ~140x versus raw accessibility-tree dumps.4 npm-