AI Debugger

AI-Powered Debugging for Every Language

AI Debugger (AIDB) brings the proven Debug Adapter Protocol (DAP) ecosystem to AI agents through a standardized Model Context Protocol (MCP) interface. Debug Python, JavaScript, TypeScript, and Java programs using the same battle-tested adapters that power VS Code—no IDE required, no heavyweight dependencies, just powerful debugging at your AI assistant's fingertips.

Read the Docs | Join Discord | Star on GitHub

Quick Install

Get started with Python debugging in under 60 seconds:

pip install ai-debugger-inc

Add to your MCP client settings (Claude Code, Cline, Cursor, etc.):

{
  "mcpServers": {
    "ai-debugger": {
      "command": "python",
      "args": ["-m", "aidb_mcp"]
    }
  }
}

Ask your AI assistant:

"Initialize debugging for Python. Debug app.py with a breakpoint at line 25."

JavaScript/Java? Visit the docs for multi-language setup.

Why AI Debugger?

Standalone & Zero Heavy Dependencies

No VS Code required. No heavyweight IDEs. Just install with pip and you're debugging––works on macOS, Linux, and Windows (WSL supported).

The core Python dependencies are lightweight and minimal:

dependencies = [
  "aiofiles",
  "mcp",
  "psutil"
]

Debug adapters are built during the release pipeline and are published as release artifacts. Once the ai-debugger-inc package is installed, your agent will use the download tool to fetch the appropriate adapter binaries automatically on first run.

Multi-Language from Day One

Debug Python, JavaScript, TypeScript, and Java with a single MCP server. AIDB is designed to support all DAP-compatible adapters, with more languages coming.

Built on the DAP Standard

AIDB uses the same Debug Adapter Protocol that powers VS Code debugging. We integrate with proven, open-source debug adapters:

• Python: debugpy (Microsoft)
• JavaScript/TypeScript: vscode-js-debug (Microsoft)
• Java: java-debug (Microsoft)

This means you get reliable, well-maintained debugging that "just works" with established patterns developers already trust.

VS Code Integration (Without VS Code)

Already have complex debug configurations in launch.json? AIDB can use them directly—making sophisticated debugging setups portable and shareable across teams without requiring VS Code installations.

Advanced Debugging Features

• Framework detection: Auto-detects pytest, jest, django, spring, flask, and more
• Conditional breakpoints: Break on user.role == "admin" or after N hits
• Logpoints: Log values without pausing execution
• Live code patching: Modify functions at runtime during debugging

Future-Ready Architecture

AIDB is built for where AI-assisted development is heading:

• CI/CD Debugging: Imagine test failures in your pipeline automatically triggering debug sessions for deeper RCA
• Agent Tooling: Native debugging capabilities for autonomous AI agents
• Cross-Platform Consistency: Same debugging API across all environments

How It Works

┌──────────────────────────────────────────────────────────────────┐
│                        Your AI Assistant                         │
│                    (Claude, GPT, Local LLMs)                     │
└────────────────────────────────┬─────────────────────────────────┘
                                 │
                                 ▼
                            MCP Protocol
┌──────────────────────────────────────────────────────────────────┐
│                      AI Debugger MCP Server                      │
│         Agent-Optimized Tools (init, step, inspect, etc.)        │
└────────────────────────────────┬─────────────────────────────────┘
                                 │
                                 ▼
                            AIDB Core API
┌──────────────────────────────────────────────────────────────────┐
│                     Debug Adapter Protocol                       │
│              Language-Agnostic Debugging Interface               │
└───────────┬────────────────────┼────────────────────┬────────────┘
            │                    │                    │
            ▼                    ▼                    ▼
    ┌───────────────┐   ┌─────────────────┐   ┌───────────────┐
    │    debugpy    │   │ vscode-js-debug │   │   java-debug  │
    │    (Python)   │   │     (JS/TS)     │   │     (Java)    │
    └───────┬───────┘   └────────┬────────┘   └───────┬───────┘
            │                    │                    │
            ▼                    ▼                    ▼
       Your Python          Your Node.js          Your Java
         Program              Program              Program

The Bridge Between AI and Proven Tools

AI Debugger acts as a translation layer, exposing the mature Debug Adapter Protocol ecosystem to AI agents through a clean, agent-optimized MCP interface. Your AI assistant gets powerful debugging capabilities, and you get the reliability of debug adapters used by millions of developers worldwide.

Learn more about the architecture →

Supported Languages

Platforms: macOS, Linux, Windows (WSL recommended; native support in progress)

Coming Soon: Built to support all DAP-compatible adapters––AIDB is designed to become the debugging standard for AI systems across every popular language and framework.

Documentation

Getting Started

• Quickstart Guide
  • Install and debug in 5 minutes
• Core Concepts
  • Sessions, breakpoints, execution flow
• Language Guides
  • Python, JavaScript, Java examples

Technical Reference

• MCP Tools Reference
  • Complete tool documentation
• API Documentation
  • Python API reference
• Advanced Workflows
  • Remote debugging, multi-session

Architecture & Design

• How It Works
  • System architecture deep dive
• DAP Protocol Guide
  • Debug Adapter Protocol reference

Development Setup

Prerequisites: Python 3.10+, Docker

Initial setup:

bash scripts/install/src/install.sh -v
./dev-cli info
./dev-cli completion install --yes  # optional

Common commands:

./dev-cli test run --coverage
./dev-cli docs serve --build-first -p 8000

Project Structure

• aidb/: Core debugging API, language adapters, session management
• aidb_mcp/: MCP server exposing debugging tools to AI agents
• aidb_cli/: Developer CLI for testing, Docker, adapter builds
• aidb_common/, aidb_logging/: Shared utilities and structured logging

For architecture details and implementation guidance, see the Developer Guide.

Robust Testing & Releases

AIDB is built with a comprehensive CI/CD pipeline:

• Thorough E2E Testing: Multi-language, multi-framework integration tests
• Automated Releases: Reliable version management and publishing
• Continuous Quality: The test suite is run nightly and on all release PRs

We catch issues early and ship features confidently, ensuring the debugging experience you depend on stays reliable.

Our entire CI/CD release pipeline executes start to finish in under 15 minutes––a target we plan to maintain.

Our Vision

Becoming the debugging standard in the MCP tools space.

As AI agents become more capable, they need debugging tools designed for their workflows—not adapted from human-centric IDEs. AIDB provides a unified, language-agnostic approach to debug any program with any AI agent through the proven MCP standard.

We're building the future of AI-assisted debugging, one DAP adapter at a time.

Contributing

We welcome contributions! See our Contributing Guide to get started.

• Contributing Guide - How to contribute
• Code of Conduct - Community standards
• Security Policy - Reporting vulnerabilities

Community & Support

• Documentation: ai-debugger.com
• Discord Community: Join the conversation
• Issues & Features: GitHub Issues

License

AI Debugger is licensed under the Apache 2.0 License. See LICENSE for details.