ZeroAlloc.NET
A 100+ package .NET ecosystem where every dispatch, registration, and validation rule is wired at compile time by Roslyn source generators — no reflection, no boxing, Native AOT safe.
Hi, I'm
I build open-source .NET libraries: ZeroAlloc for performance, and AI libraries for C# such as Minos.NET, Rag.NET and AI.Sentinel.
A 100+ package .NET ecosystem where every dispatch, registration, and validation rule is wired at compile time by Roslyn source generators — no reflection, no boxing, Native AOT safe.
Roslyn-based MCP server giving AI agents semantic understanding of .NET codebases — 67 tools for navigation, call graphs, diagnostics and code fixes, safe refactoring, and code-quality auditing.
MCP server for .NET memory profiling with AI-actionable code fix suggestions, powered by JetBrains dotMemory.
Unofficial .NET client for TypeSafe's Jev decision models — declare typed questions as a C# type, get calibrated answers back. Source-generated and Native AOT safe.
Security monitoring middleware for IChatClient — 56 detectors for prompt injection, hallucination, PII leakage, and operational anomalies, with an intervention engine and audit forwarders.
A modular RAG pipeline library for .NET, built on the Microsoft.Extensions.AI abstractions.
An unofficial .NET client for TypeSafe's Jev decision models: typed questions declared as a C# type, calibrated answers back, source-generated and Native AOT safe.
A ZeroAlloc-native agent framework for .NET on Microsoft Agent Framework, where tool authorization, curated memory, and model-boundary scanning are runtime ports rather than afterthoughts.
A .NET 10 application for AI-driven task execution — two front ends over one CQRS stack, failures as values, and a Thalos agent layer, orchestrated with Aspire.