dotnet-ai-kit

The AI brain for .NET — 124 skills, 13 agents, one command to ship features

124 skills · 13 agents · 27 commands · 16 rules · 12 profiles · 16 knowledge docs · 7 hooks · 11 CLI commands · 4 AI hosts

---

The Problem

When you use AI coding assistants on .NET projects, the AI doesn't know your architecture. It generates layered code in your VSA project, anemic models in your DDD project, and everything in one repo for your CQRS microservices. You spend hours fixing AI output to match your patterns.

dotnet-ai-kit fixes this. It gives AI deep .NET intelligence — your architecture, your conventions, your lifecycle — so every line of generated code fits your project.

---

See It in Action

_{/dai.do running the full 9-phase lifecycle inside Claude Code}

---

Quick Start

CLI Installation

# Install the CLI
uv tool install dotnet-ai-kit --from git+https://github.com/FaysilAlshareef/dotnet-ai-kit.git

# Check version
dotnet-ai --version

# Initialize your project (auto-detects architecture)
dotnet-ai init . --ai claude

# Initialize with permissions applied in one step
dotnet-ai init . --ai claude --permissions standard

# Preview everything before it runs
dotnet-ai init . --ai claude --dry-run

# Build a complete feature with one command
/dai.do "Add order management with tracking"

# That's it → spec → plan → code → test → review → PR

See installation demo

Claude Code Plugin (Recommended)

# Add the marketplace
/plugin marketplace add FaysilAlshareef/dotnet-ai-kit

# Install
/plugin install dotnet-ai-kit

All 27 commands, 124 skills, 13 agents, 16 rules, and 7 hooks are available immediately.

See plugin install demo

Required dependencies

dotnet-ai init auto-installs both dependencies when they are missing (skip with --no-install-deps).

csharp-ls — C# language server

Provides semantic symbol/reference navigation. Declared as a csharp-lsp plugin dependency in the Claude plugin manifest — not bundled in .mcp.json.

# Manual install (if auto-install is skipped)
dotnet tool install -g csharp-ls

codebase-memory-mcp ≥ 0.6.1 — project graph + memory

Registered in .mcp.json. Provides lazy project-graph queries instead of broad reads.

# Cross-platform via PyPI (auto-installed by dotnet-ai init)
pip install "codebase-memory-mcp>=0.6.1"

MCP health is recorded in .dotnet-ai-kit/mcp-state.yml (accepted / below-minimum / unavailable) after each init or configure run.

Without codebase-memory-mcp, project-graph questions fall back to csharp-ls + grep/read. Without csharp-ls, the AI uses grep-based analysis which works but uses significantly more context tokens on large codebases.

Claude Code compatibility

Claude Code v2.1.85+ recommended. v1.0 is supported with reduced hook fidelity — the dynamic architecture hook falls back to a command-pattern matcher (no per-tool if: scoping).

---

Security & Permissions

What the Plugin Accesses