Skip to main content
Glama

agentic-debugger

An MCP (Model Context Protocol) server that enables interactive debugging with code instrumentation for AI coding assistants. Inspired by Cursor's debug mode.

Works with any MCP-compatible AI coding tool:

  • Claude Code

  • Cursor

  • Windsurf

  • Cline

  • GitHub Copilot

  • Kiro

  • Zed

  • And more...

Features

  • Live code instrumentation - Inject debug logging at specific lines

  • Variable capture - Log variable values at runtime

  • Multi-language support - JavaScript, TypeScript, and Python

  • Browser support - CORS-enabled for browser JS debugging

  • Clean removal - Region markers ensure instruments are fully removed

Related MCP server: VSCode Debug MCP

Installation

Add to your MCP configuration:

{
  "mcpServers": {
    "debug": {
      "command": "npx",
      "args": ["-y", "agentic-debugger"]
    }
  }
}

Configuration file locations:

  • Claude Code: ~/.mcp.json

  • Cursor: .cursor/mcp.json in your project or ~/.cursor/mcp.json

  • Other tools: Check your tool's MCP documentation

Global install

npm install -g agentic-debugger

Then configure:

{
  "mcpServers": {
    "debug": {
      "command": "agentic-debugger"
    }
  }
}

Available Tools

Tool

Description

start_debug_session

Start HTTP server for log collection

stop_debug_session

Stop server and cleanup

add_instrument

Insert logging code at file:line

remove_instruments

Remove debug code from file(s)

list_instruments

Show all active instruments

read_debug_logs

Read captured log data

clear_debug_logs

Clear the log file

How It Works

  1. Start session - Spawns a local HTTP server (default port 9876)

  2. Add instruments - Injects fetch() calls that POST to the server

  3. Reproduce bug - Run your code, instruments capture variable values

  4. Analyze logs - Read the captured data to identify issues

  5. Cleanup - Remove all instruments and stop the server

Debug Workflow Example

You: "Help me debug why the total is NaN"

AI Assistant:
1. Starts debug session
2. Reads your code to understand the logic
3. Adds instruments at suspicious locations
4. "Please run your code to reproduce the issue"

You: *runs code* "Done"

AI Assistant:
5. Reads debug logs
6. "I see `discount` is undefined at line 15..."
7. Removes instruments
8. Fixes the bug
9. Stops debug session

Instrument Examples

JavaScript/TypeScript

// #region agentic-debug-abc123
fetch('http://localhost:9876/log', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({
    id: 'abc123',
    location: 'cart.js:15',
    timestamp: Date.now(),
    data: { total, discount, items }
  })
}).catch(() => {});
// #endregion agentic-debug-abc123

Python

# region agentic-debug-abc123
try:
    import urllib.request as __req, json as __json
    __req.urlopen(__req.Request(
        'http://localhost:9876/log',
        data=__json.dumps({
            'id': 'abc123',
            'location': 'cart.py:15',
            'timestamp': __import__('time').time(),
            'data': {'total': total, 'discount': discount}
        }).encode(),
        headers={'Content-Type': 'application/json'}
    ))
except: pass
# endregion agentic-debug-abc123

Supported Languages

Language

Extensions

JavaScript

.js, .mjs, .cjs

TypeScript

.ts, .tsx

Python

.py

Requirements

  • Node.js >= 18.0.0

  • An MCP-compatible AI coding assistant

License

MIT

Available Tools

7 tools
add_instrumentB

Add a debug instrument at a specific line in a file. The instrument will log variable values when executed.

Inputs: file (required), line (required), capture.

clear_debug_logsB

Clear all collected debug logs.

Inputs: no named parameters.

list_instrumentsB

List all active debug instruments.

Inputs: no named parameters.

read_debug_logsB

Read the collected debug logs from the current session.

Inputs: format.

remove_instrumentsC

Remove debug instruments from files.

Inputs: file.

start_debug_sessionC

Start a debug session. This starts a local HTTP server to receive logs from instrumented code.

Inputs: port.

stop_debug_sessionA

Stop the current debug session and shut down the log server.

Inputs: no named parameters.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.7/5.0

    Scored across 7 tools

    Disambiguation5/5

    Each tool has a clearly distinct purpose with no overlap: adding, listing, and removing instruments; starting and stopping debug sessions; and reading and clearing logs. The descriptions make it easy to differentiate between them, such as distinguishing 'add_instrument' from 'remove_instruments' and 'read_debug_logs' from 'clear_debug_logs'.

    Naming Consistency5/5

    All tool names follow a consistent verb_noun pattern using snake_case, such as 'add_instrument', 'clear_debug_logs', and 'start_debug_session'. There are no deviations in naming conventions, making the set predictable and easy to understand.

    Tool Count5/5

    With 7 tools, the count is well-scoped for a debugger server, covering core operations like instrument management, session control, and log handling. Each tool earns its place without feeling excessive or insufficient for the domain.

    Completeness5/5

    The tool set provides complete coverage for a debugger's lifecycle: it includes CRUD operations for instruments (add, list, remove), session management (start, stop), and log handling (read, clear). There are no obvious gaps, as all essential workflows are supported without dead ends.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers