Debugger MCP
Enables AI agents like Claude to control VS Code's debugger, supporting any language with a Debug Adapter Protocol implementation.
README
Debugger MCP
A VS Code extension that exposes an MCP (Model Context Protocol) server, allowing AI agents like Claude Code and Claude Desktop to control VS Code's debugger. Language-agnostic — works with any debug adapter (Node.js, Python, C++, Go, Rust, Java, etc.).
Features
- 22 MCP tools for full debugger control
- Works with any debugger that implements the Debug Adapter Protocol
- Zero-config for Claude Code — auto-updates
~/.claude.jsonon activation - Execution commands block until the debugger stops and return the stop location + local variables in a single response
- Captures program output (stdout/stderr) for inspection
Architecture
Claude Code / Desktop / any MCP client
| (HTTP: Streamable HTTP transport)
v
VS Code Extension (MCP server on localhost:<port>/mcp)
| (vscode.debug.* API)
v
Debug Adapter Protocol → Any Debugger
The extension hosts the MCP server directly inside the VS Code process — no separate server or sidecar needed. On activation it starts a localhost HTTP server (default port 45557) and automatically writes the connection URL to ~/.claude.json.
Installation
VS Code Marketplace
Install from the VS Code Marketplace, or search for "Debugger MCP" in the Extensions view (Ctrl+Shift+X / Cmd+Shift+X).
Or from the command line:
code --install-extension speakeasy.debugger-mcp
The extension activates automatically on startup — no manual setup required. It starts the MCP server and auto-configures Claude Code.
From source (development)
git clone git@github.com:speakeasy-api/debugger-mcp.git
cd debugger-mcp
npm install
npm run compile
Then press F5 in VS Code to launch the Extension Development Host.
MCP Client Configuration
Claude Code (zero-config)
The extension automatically registers itself in ~/.claude.json on activation. No manual setup needed — just install the extension, reload VS Code, and start using Claude Code.
To verify it's configured, run:
claude mcp list
You should see vscode-debugger listed as an HTTP server.
If you prefer to configure manually, or if auto-config is disabled:
claude mcp add --transport http vscode-debugger http://localhost:45557/mcp
Claude Desktop
Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):
{
"mcpServers": {
"vscode-debugger": {
"type": "http",
"url": "http://localhost:45557/mcp"
}
}
}
VS Code (Copilot / other MCP clients)
Add to .vscode/mcp.json in your workspace:
{
"servers": {
"vscode-debugger": {
"type": "http",
"url": "http://localhost:45557/mcp"
}
}
}
Alternative: stdio transport
A standalone MCP server process is also available for clients that only support stdio transport. See the extension's dist/mcp-server.js — it discovers the running extension via a port file and proxies MCP requests over HTTP.
Settings
| Setting | Default | Description |
|---|---|---|
debuggerMcp.port |
45557 |
Preferred port for the MCP server. Falls back to a random port if taken. |
debuggerMcp.autoConfigureClaude |
true |
Automatically update ~/.claude.json with the MCP server URL on activation. |
Tools
Session Management
| Tool | Description |
|---|---|
debug_start |
Start a debug session from a launch.json config name or inline configuration |
debug_stop |
Stop a debug session |
debug_restart |
Restart a debug session |
debug_list_sessions |
List active debug sessions |
debug_list_configurations |
List available launch.json configurations |
Execution Control
| Tool | Description |
|---|---|
debug_continue |
Resume execution until next breakpoint or exit. Returns stop location + locals |
debug_pause |
Pause a running program. Returns stop location + locals |
debug_step_over |
Step over (next line). Supports count for batch stepping |
debug_step_into |
Step into function calls. Supports count for batch stepping |
debug_step_out |
Step out of current function. Supports count for batch stepping |
Breakpoints
| Tool | Description |
|---|---|
debug_set_breakpoint |
Set a source breakpoint (supports condition, hit count, logpoint) |
debug_set_function_breakpoint |
Set a function-name breakpoint |
debug_remove_breakpoint |
Remove by ID, file+line, all in file, or all |
debug_list_breakpoints |
List all breakpoints with details |
debug_toggle_breakpoint |
Enable/disable a breakpoint without removing it |
State Inspection
| Tool | Description |
|---|---|
debug_get_threads |
List threads |
debug_get_stack_trace |
Get call stack for a thread |
debug_get_scopes |
Get variable scopes for a stack frame |
debug_get_variables |
Get variables in a scope or expand a structured variable |
debug_evaluate |
Evaluate an expression in the debug context |
Status & Output
| Tool | Description |
|---|---|
debug_status |
Quick "where am I?" — session, stop location, and locals |
debug_get_output |
Get captured program output (stdout/stderr) |
Build Scripts
| Script | Description |
|---|---|
npm run compile |
Type-check and build both bundles |
npm run build |
Production build (minified, no source maps) |
npm run watch |
Watch mode for development |
npm run package |
Production build + package into .vsix |
npm run check-types |
Type-check only (no emit) |
How It Works
- Extension activates on VS Code startup and starts an HTTP server on
localhost:45557 - MCP endpoint at
/mcpuses the Streamable HTTP transport protocol — each client gets its own session - Auto-config writes the server URL to
~/.claude.jsonso Claude Code discovers it automatically - Debug commands are forwarded to the VS Code debug API via an in-process
DirectClient - DAP message interception captures stopped events, program output, and exit codes in real-time
- Execution commands (continue, step) block until the debugger pauses, then automatically gather the stop location and local variables
Security
- HTTP server binds to
127.0.0.1only (no network exposure) - MCP endpoint uses the standard Streamable HTTP transport protocol
- Port file (for legacy stdio transport) includes random auth token and PID for stale-file detection
License
MIT
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.