The hands-on Model Context Protocol (MCP) workshop from DevCon 2026 now ships with a complete Python track, joining the existing Node.js and .NET content. If you build in Python, you can now work through the entire workshop end-to-end without leaving your language of choice — the code, the tooling, and the docs are all there.
What is the MCP Workshop?
The workshop is a guided, seven-step journey that takes you from a bare MCP server all the way to a production-ready agent grounded in real Autodesk Platform Services (APS) APIs:
- Build an MCP server and connect it to VS Code and GitHub Copilot.
- Extend it with external tools — a filesystem tool and a live weather API.
- Connect to the Autodesk Platform Services (APS) API using 2-legged OAuth to list and create OSS buckets.
- Add 3-legged OAuth so your agent can call APS APIs on behalf of a signed-in user.
- Build your own MCP client and wire in an LLM to create a fully autonomous agent.
Every step is now available in Python, right alongside the Node.js and .NET versions.
What's in the Python Track?
The Python track is purely additive — nothing in the existing Node.js or .NET experience changed. What's new:
- Language-tabbed docs. Every chapter now has switchable Node.js, .NET, and Python code tabs, so you can follow whichever runtime you prefer with a single click.
- Seven Python code snapshots. A complete `-python` code state for each stage of the workshop, from the first `add` and `greet` tools to the full APS-connected agent — so you can jump in at any step.
- The official MCP Python SDK. Servers are built with FastMCP, declaring tools through clean decorators and Pydantic-typed parameters, served over streamable HTTP — the same patterns you'd use in production.
- Full APS integration in Python. 2-legged and 3-legged OAuth, OSS bucket operations, a custom MCP client, and an interactive agent with an AEC-focused system prompt — all in idiomatic Python.
Why This Matters
- Meet developers where they are. Python is the lingua franca of AI and automation. Now that community can learn MCP in the language they already reach for.
- Real APS scenarios. The workshop connects to live APS APIs, so what you learn maps directly to the applications you ship.
- Low-boilerplate, high-clarity. FastMCP and Pydantic let you express a tool server in a handful of readable lines, so the protocol — not the plumbing — stays front and center.
- One workshop, three runtimes. Teams that mix Node.js, .NET, and Python can standardise on a single curriculum and let each developer pick their track.
Where to Find It
The complete source code is available on GitHub: autodesk-platform-services/mcp-devcon2026
Whether you're a seasoned Python developer or just curious how MCP fits into the Python world, there's never been a better time to build your first agent on Autodesk Platform Services.