vscode-playwright-mcp
Gives LLMs full control over VS Code — click UI elements, type code, run commands, take screenshots, read editor state, and record GIFs.
README
vscode-playwright-mcp
A VS Code-aware layer on top of @playwright/mcp. Gives LLMs full control over VS Code — click UI elements, type code, run commands, take screenshots, read editor state, and record GIFs.
Use cases:
- E2E testing of VS Code extensions
- Automated VS Code workflows and tasks
- AI-driven code editing and refactoring
- Interactive demos and tutorials
- Testing themes, keybindings, and settings
How It Works
Built on top of @playwright/mcp, which provides generic Electron automation (click, type, screenshot, etc.). We add a VS Code-specific layer: isolated launch with temp user-data-dir, command palette automation, editor state scraping, and Monaco-aware tools.
Generic tools from @playwright/mcp are aliased to vscode_* names for a unified LLM experience. Before vscode_launch, only 1 tool is visible (lazy discovery saves ~10k context tokens). After launch, all 26 tools appear.
Requirements
- Node.js >= 22
- VS Code installed locally (auto-detected on macOS and Linux)
Quick Start
{
"mcpServers": {
"vscode": {
"command": "npx",
"args": ["-y", "vscode-playwright-mcp"]
}
}
}
Custom VS Code binary:
{
"mcpServers": {
"vscode": {
"command": "npx",
"args": ["-y", "vscode-playwright-mcp", "--vscode-path", "/path/to/code"]
}
}
}
Tools
Lifecycle
| Tool | Description |
|---|---|
vscode_launch |
Launch an isolated VS Code instance. Options: workspace, extensions, viewport size, settings. |
vscode_close |
Close VS Code and clean up temp files/processes. |
Vision (aliased from @playwright/mcp)
| Tool | Description |
|---|---|
vscode_screenshot |
Full-window screenshot. |
vscode_snapshot |
Accessibility tree with [ref=eN] annotations on interactive elements. |
vscode_zoom |
Cropped screenshot of a specific region (native — Monaco-aware). |
vscode_find_element |
Search a11y tree by role/name, returns refs (native). |
Mouse & Keyboard (aliased from @playwright/mcp)
| Tool | Description |
|---|---|
vscode_click |
Click by ref, element description, or coordinates. |
vscode_type |
Type text into focused element. |
vscode_press_key |
Press key combinations (e.g., Control+Shift+P). |
vscode_hover |
Hover to reveal tooltips. |
vscode_drag |
Drag from one point to another. |
vscode_click_xy |
Click at exact pixel coordinates. |
vscode_hover_xy |
Hover at exact pixel coordinates. |
vscode_drag_xy |
Drag between exact pixel coordinates. |
vscode_select_option |
Select from dropdown/combobox. |
vscode_fill_form |
Fill form inputs. |
VS Code Commands (native)
| Tool | Description |
|---|---|
vscode_run_command |
Execute a VS Code command via Command Palette automation. |
vscode_ensure_file |
Open and activate a file by path with verification. |
State (native)
| Tool | Description |
|---|---|
vscode_get_state |
Read editor state: active file, cursor, diagnostics, completions, peek widget. |
vscode_get_hover |
Read hover tooltip content as text. |
Advanced (mixed)
| Tool | Description |
|---|---|
vscode_evaluate |
Run JavaScript in the VS Code renderer process (aliased). |
vscode_wait_for |
Wait for selector state, text, or delay (aliased). |
vscode_console |
Retrieve console messages from renderer (aliased). |
vscode_resize |
Resize the VS Code window and viewport (native). |
vscode_scroll |
Scroll up/down by amount (native). |
vscode_gif |
Record actions as animated GIF (native). |
Interaction Strategies
-
Ref-based (preferred for UI):
vscode_snapshot→ see[ref=eN]annotations →vscode_click(ref="e5"). Deterministic, no coordinate guessing. -
Keyboard-first (preferred for navigation):
vscode_snapshot→ see shortcuts →vscode_press_key. Fastest for commands. -
Visual (required for editor content): Monaco editor is opaque to accessibility APIs.
vscode_screenshotorvscode_zoom→ read the image →vscode_click_xy(x, y).
How is this different from Playwright MCP?
This project is built on top of @playwright/mcp, not a replacement. Playwright MCP provides generic Electron automation. We add:
- Isolated VS Code launch with temp user-data-dir, suppressed Welcome tab, settings injection
- Command palette automation (
vscode_run_command) — execute commands by name - Editor state scraping (
vscode_get_state) — cursor, diagnostics, completions, peek widget - Monaco-aware tools (
vscode_zoom,vscode_find_element) — the editor is invisible to generic a11y tools - Lazy tool discovery — 1 tool visible before launch, all ~26 after (saves context tokens)
- GIF recording of VS Code sessions
Development
npm run build # Build with tsup
npm run dev # Build in watch mode
npm test # Run unit tests (vitest)
npm run test:integration # Run integration tests (requires VS Code)
npm run test:e2e # Run E2E tests (full MCP protocol loop)
npm run typecheck # Type-check with tsc
npm run lint # Lint with ESLint
License
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.