WinAiDbg MCP
AI-powered Windows crash dump analysis platform that provides structured access to Microsoft debugging tools through the Model Context Protocol, making complex crash investigation accessible to AI systems
WinAiDbg
AI-Powered Windows Crash Dump Analysis Platform

WinAiDbg is a comprehensive Model Context Protocol (MCP) server that provides AI systems with advanced Windows crash dump analysis capabilities. It combines the power of Microsoft's debugging tools (WinDBG/CDB) with intelligent analysis workflows, making professional-grade crash investigation accessible to AI assistants.
๐ฏ What is WinAiDbg?
WinAiDbg is a platform that provides structured access to Windows debugging tools through the Model Context Protocol. It makes crash dump analysis more accessible by providing standardized tools and real-time progress tracking.
The Problem We Solve
Traditional Crash Analysis Challenges:
- ๐ Complexity: Requires deep knowledge of Windows internals and debugging tools
- โฑ๏ธ Time-consuming: Manual analysis can take hours or days
- ๐ง Expertise Required: Need specialized debugging skills and experience
- ๐ Inconsistent Results: Different analysts may reach different conclusions
- ๐ง Tool Complexity: WinDBG/CDB have steep learning curves
WinAiDbg Solution:
- ๐ง Structured Access: Provides standardized tools for debugging operations
- โก Real-time Updates: Live progress tracking and notifications
- ๐ Consistent Results: Provides structured output formats
- ๐ MCP Integration: Works with AI clients through Model Context Protocol
- ๐ ๏ธ Professional Tools: Built on Microsoft's industry-standard debugging infrastructure
โจ Key Features
- ๐ Advanced Crash Analysis: Leverage WinDBG/CDB for comprehensive dump analysis
- ๐ค AI-Native Design: Built specifically for AI agent integration via MCP
- โก Command Batching: Intelligent command grouping for improved throughput
- ๐ Real-time Notifications: Live updates during analysis operations
- ๐ก๏ธ Session Management: Robust session lifecycle with automatic cleanup
- ๐ฏ Extensible Architecture: PowerShell-based extension system for custom workflows
- ๐ Structured Results: Parse debugging output into AI-friendly formats
๐ฏ Quick Start
Prerequisites
- .NET 8.0 Runtime or SDK
- Windows Debugging Tools (WinDBG/CDB) - Download from Microsoft
Installation
For local development (or a one-off manual run), clone the repo, build it, and run WinAiDbg as a normal console process.
This approach is typically the best fit for STDIO-based integrations, where the client starts WinAiDbg and communicates over stdin/stdout.
# Clone the repository
git clone https://github.com/CapulusCodeNinja/mcp-win-ai-dbg.git
cd mcp-win-ai-dbg
# Build the project
dotnet build
# Run the server (interactive / foreground)
dotnet run --project winaidbg/winaidbg.csproj
๐ก๏ธ Service Mode
Use Service Mode when you want WinAiDbg to run in the background (always-on) and be managed by the Windows Service Control Manager.
This is the recommended setup for production/long-running environments.
This approach is typically the best fit for HTTP-based integrations, where clients connect to a long-running WinAiDbg instance over the network.
# Install WinAiDbg as a Windows Service
dotnet run --project winaidbg/winaidbg.csproj -- service install
# Start the service
net start WinAiDbg
# Stop the service
net stop WinAiDbg
# Uninstall the Windows Service
dotnet run --project winaidbg/winaidbg.csproj -- service uninstall
๐ค AI Integration
This section is an index of supported AI development environments. Open the relevant integration page below for the environment-specific setup and usage instructions (configuration files, transport selection for STDIO vs HTTP, and run/debug tips).
In similar fashion, you can adapt these configurations for other MCP-compatible clients.
๐ ๏ธ Available MCP Tools
Core Analysis Tools
winaidbg_open_dump_analyze_session- Open a crash dump for analysiswinaidbg_enqueue_async_dump_analyze_command- Execute WinDBG commands asynchronouslywinaidbg_get_dump_analyze_commands_status- Get status of all commands (bulk polling)winaidbg_read_dump_analyze_command_result- Read individual command resultswinaidbg_cancel_dump_analyze_command- Cancel a running commandwinaidbg_close_dump_analyze_session- Close session and cleanup resources
MCP Resources
sessions- List all active debugging sessionscommands- List commands with filtering by session/stateextensions- List available PowerShell extension scriptsusage- Complete usage guide with exampleshealth- System health and metrics
๐ Example Workflow
Typical AI-driven crash analysis workflow:
- Open Session:
winaidbg_open_dump_analyze_sessionwith dump file path - Queue Commands: Use
winaidbg_enqueue_async_dump_analyze_commandfor:!analyze -v(automatic crash analysis)kL(stack trace with source lines)!threads(thread information)
- Monitor Progress:
winaidbg_get_dump_analyze_commands_status(bulk polling) - Retrieve Results:
winaidbg_read_dump_analyze_command_resultfor each command - Close Session:
winaidbg_close_dump_analyze_sessionfor cleanup
Example conversation with AI assistant:
User: "Analyze the crash dump at C:\dumps\myapp.dmp"
AI: I'll analyze that crash dump for you.
[Opens session, queues !analyze -v, kL, !threads commands]
Analysis Results:
- Exception: Access Violation (0xC0000005)
- Faulting Module: myapp.exe
- Root Cause: Null pointer dereference in ProcessData()
- Stack shows: ProcessData+0x42 called with null buffer
Recommendation: Add null check before buffer access.
๐๏ธ Architecture
Modular Structure
winaidbg/ - Main application (entry point, hosting)
winaidbg_config/ - Configuration and logging
winaidbg_engine/ - Debug engine + subcomponents (CDB sessions, queue, batching, extensions)
winaidbg_protocol/ - MCP protocol layer (tools, resources)
winaidbg_setup/ - Service installation and management
winaidbg_external_apis/ - Shared utilities (file system, process, etc.)
winaidbg_web/ - Static admin UI and docs
Key Design Patterns
- Modular Architecture: Clear separation of concerns across libraries
- Singleton Pattern: Core engine accessible without DI overhead
- Command Queue: Asynchronous command processing with state management
- Batching: Transparent command grouping for improved performance
- Event-Driven: Real-time notifications for command state changes
๐งช Testing & Quality
Test Statistics
- 1265 total tests (all passing)
- 85.7% line coverage (target: 75%)
- 75.6% branch coverage (target: 75%)
- Fast execution (~16 seconds for full suite)
- Zero build warnings
โ๏ธ Configuration
Configuration is documented in these section pages:
- Logging: Logging.md
- WinAiDbg.Server: Server.md
- WinAiDbg.Transport: Transport.md
- WinAiDbg.Debugging: Debugging.md
- WinAiDbg.Validation: Validation.md
- WinAiDbg.AutomatedRecovery: AutomatedRecovery.md
- WinAiDbg.Service: Service.md
- WinAiDbg.SessionManagement: SessionManagement.md
- WinAiDbg.Extensions: Extensions.md
- WinAiDbg.Batching: CommandBatching.md
- WinAiDbg.ProcessStatistics: ProcessStatistics.md
- IpRateLimiting: IpRateLimiting.md
๐ Features
Features are documented in the pages below:
- Extension system: ExtensionSystem.md
- Real-time notifications: RealTimeNotifications.md
- Advanced crash analysis: AdvancedCrashAnalysis.md
- AI-native design: AiNativeDesign.md
- Command batching: CommandBatching.md
- Session management: SessionManagement.md
- Structured results: StructuredResults.md
๐ License
This project is licensed under the Apache License 2.0 - see the LICENSE file for details.
The Apache License 2.0 allows you to:
- โ Use the software commercially
- โ Modify and distribute
- โ Sublicense
- โ Use patent claims
- โ ๏ธ Include copyright notice
๐ Acknowledgments
- Model Context Protocol - MCP specification
- C# SDK for MCP - MCP implementation
- Microsoft Debugging Tools - WinDBG/CDB
- NLog - Logging framework
Related Servers
Scout Monitoring MCP
sponsorPut performance and error data directly in the hands of your AI assistant.
Alpha Vantage MCP Server
sponsorAccess financial market data: realtime & historical stock, ETF, options, forex, crypto, commodities, fundamentals, technical indicators, & more
Octomind
Create and manage end-to-end tests using the Octomind platform.
Memory Bank MCP
An AI-assisted development plugin that maintains persistent project context using structured markdown files for goals, decisions, and progress.
Apple HIG
Provides instant access to Apple's Human Interface Guidelines, with content auto-updated periodically.
Remote MCP Server (Authless)
A template for deploying a remote, auth-less MCP server on Cloudflare Workers.
Bonsai MCP
An MCP server that integrates IFC model support using Bonsai BIM (Blender) and IfcOpenShell.
Email MCP Server by Sidemail
Let AI agents write & manage your SaaS emails
MCPAgent
An intelligent agent framework based on MCP, supporting multiple large language models and tool integrations for testing single-agent effectiveness.
Juspay MCP Tools
Interact with Juspay APIs for payment processing and merchant dashboard management.
Command-Line MCP Server
A secure MCP server for executing terminal commands with controlled directory access and command permissions.
MCP Documentation Server
An AI-powered documentation server for code improvement and management, with Claude and Brave Search integration.