GitHub

Daniel Meppiel

I work on the methodology and tools for building software with AI agents at enterprise scale.

I'm a Software Global Black Belt at Microsoft and the creator of APM (Agent Package Manager) — the open-source tool teams use to manage AI agent configurations the way they manage code. I designed the PROSE framework for writing agent specifications that hold up in production, and I wrote The Agentic SDLC Handbook to put the whole methodology in one place.

Before Microsoft: GitHub, Sonar, four years as CTO and co-founder of WeGaw, and earlier engineering work at CERN.


🔭 What I'm working on now

  • 📘 The Agentic SDLC Handbook — released as a pre-release edition; shipping refinements and new case studies continuously.
  • 📦 microsoft/apm — Agent Package Manager. apm install, apm compile. The 2-command pattern Fortune 500 teams are adopting.
  • 📐 microsoft/agentrcGet your repo ready for AI. The companion convention for instrumenting codebases.
  • 🧬 genesis — an architect's discipline for designing agentic systems across primitives (skills, agents, instructions, orchestrators, plans). Markdown that steers an LLM is code — design it before you write it. Portable across Claude Code, Copilot, Cursor, OpenCode, Codex.
  • ✍️ The AI-Native Mind — short essays on what AI-native development looks like once the demo ends.

Updated monthly. If this is stale, I owe you a coffee.


🧰 Pinned work — the throughline

Three tools → the methodology → proof of range.

Project Why it's here
microsoft/apm Agent Package Manager. How AI agent capabilities are packaged, versioned, and shared across teams.
microsoft/agentrc The convention for getting a codebase ready to be worked on by AI agents.
genesis An architect's discipline for designing agentic systems — across skills, agents, instructions, orchestrators, plans. Harness-portable.
agentic-sdlc-handbook The book that ties all three together — written using the methodology it teaches.
awesome-ai-native The PROSE framework, distilled. Start here if you're a practitioner.
meeting-2-code End-to-end proof: Teams meeting → shipped Azure feature, fully agentic.
oracle-mcp-server MCP server for very large Oracle schemas — serious engineering outside the AI-native bubble.

🎙️ Speaking & writing

I talk about what changes — for the org, the codebase, and the developer's day — when AI coding agents become a real part of how software gets built.


A note on the journey

I don't think anyone has this fully figured out yet — including me. APM, agentrc, and genesis were born from ingenuity and breaking things, and they're evolving fast. If you're building in this space, I'd genuinely love to compare notes.

→ The fastest way to stay in the loop: subscribe to the handbook updates.

Opinions here are my own and don't represent my employer.

Read the original on github.com ↗