AI-assisted debugging suite for .NET. A single MCP server that provides two complementary tools for AI agents and a live dashboard for us humans. Still testing it out in my own every day workflows.
Interactive debugging over the Debug Adapter Protocol. Launch .NET projects, set breakpoints, step through execution, inspect locals and call stacks, evaluate watches, and diagnose exceptions.
AOP method tracing via DI interception. The OrangeCandy.Observe NuGet package is added to the target application. Every method call on DI-registered interfaces is captured (arguments, return values, timing, exceptions) and streamed to the debug server in real time.
| Path | Description |
|---|---|
mcp-server/ |
Go MCP server. Binary includes embedded web UI. |
ui/ |
Svelte dashboard served at http://localhost:9119. |
sdk/OrangeCandy.Observe/ |
.NET NuGet package for AOP observation. |
test-app/ |
Example .NET application demonstrating both tools. |
Download the binary for your platform from the releases page and place it on your PATH.
Add to your MCP client configuration (e.g. ~/.yourBrandOfRobot.json):
{
"mcpServers": {
"orangecandy-debug": {
"command": "/path/to/orangecandy-debug"
}
}
}The server also requires netcoredbg on the PATH for DAP support.
See sdk/OrangeCandy.Observe/README.md.
The MCP server embeds a web dashboard at http://localhost:9119. It connects via WebSocket and shows the live debug session: timeline of events, source context, locals, watches, call stack, and method call flow.
- .NET 10 (for the SDK and target applications)
- Go 1.26+ (to build the MCP server from source)
- Node 24+ and pnpm 10+ (to build the UI from source)
- netcoredbg on PATH
Apache 2.0 with Commons Clause. Free to use, modify, and redistribute. You may not sell the software or a product/service whose value derives substantially from it. See LICENSE.