An enterprise-grade AI automation platform built at the intersection of autonomous agents, prompt governance, and multi-agent coordination.
Three production-quality systems — written in C#/.NET 10, TypeScript, and Python — that demonstrate how AI agents should be built for real-world reliability, compliance, and scale.
graph TD
User["Developer / Operator"] -->|"GitHub Issue"| GSD
User -->|"multi-agent run"| AG
subgraph "Coding-Autopilot-System Portfolio"
subgraph "Layer 3 — Multi-Agent Coordination"
AG["autogen (Python)<br/>Microsoft Agent Framework<br/>Gemini/Claude fallback<br/>AG-UI Command Center"]
end
subgraph "Layer 2 — Autonomous Workflow Engine"
GSD["gsd-orchestrator (C#/.NET 10)<br/>Autonomous GitHub Agent<br/>State Machine + Polly Resilience<br/>JSON-RPC MCP Client"]
end
subgraph "Layer 1 — Prompt Governance"
PI["Promptimprover (TypeScript)<br/>MCP Server Middleware<br/>RAG Neural Snippets<br/>ISO 27001 Compliance"]
end
AG -->|"delegates governed prompts via MCP"| GSD
GSD -->|"prompt refinement + compliance"| PI
AG -.->|"optional direct governance"| PI
end
subgraph "External Systems"
GH["GitHub API<br/>(Issues, PRs, Branches)"]
CLAUDE["Anthropic Claude API"]
GEMINI["Google Gemini API"]
end
GSD -->|"reads issues, creates PRs"| GH
GSD -->|"plans code changes"| CLAUDE
AG -->|"multi-model routing"| CLAUDE
AG -->|"fallback model"| GEMINI
PI -->|"serves MCP protocol"| GSD
PI -->|"serves MCP protocol"| AG
gsd-orchestrator — Autonomous GitHub Agent
C# / .NET 10 — Reads GitHub issues and autonomously plans, branches, edits, and opens PRs using Claude AI. Implements a state machine with Polly resilience, file checkpointing for durability, and a JSON-RPC MCP stdio client for prompt governance integration.
Enterprise patterns: State machine, dependency injection, Polly resilience policies, structured logging, async/await throughout with CancellationToken propagation.
Promptimprover — Prompt Governance MCP Server
TypeScript — MCP server middleware implementing prompt governance as a first-class infrastructure concern. RAG-based neural snippet retrieval, compounding memory, auto-heal middleware, and ISO 27001 compliance framing.
Enterprise patterns: MCP protocol server, RAG architecture, middleware pipeline, compliance-first design.
autogen — Multi-Agent Coordination
Python — Multi-agent automation built on Microsoft AutoGen with model-fallback resilience (Anthropic Claude / Google Gemini), AG-UI Command Center for agent state observability, and DevUI integration for operator-in-the-loop control.
Enterprise patterns: Agent framework integration, model-fallback routing, observability tooling, operator control plane.
| Area | Technologies |
|---|---|
| Languages | C# / .NET 10 · TypeScript · Python |
| AI Providers | Anthropic Claude · Google Gemini |
| Protocols | Model Context Protocol (MCP) · JSON-RPC 2.0 |
| Patterns | State machine · RAG · Multi-agent coordination |
| Resilience | Polly retry/circuit-breaker · Model fallback routing |
| Infrastructure | GitHub Actions · GitHub API · GitHub MCP Server |
| Compliance | ISO 27001 framing |
Built by @OgeonX-Ai — AI Engineer and Senior .NET Developer