# I received the AAIF Outstanding Technical Contribution Award > Source: > Published: 2026-10-27 09:00:00+00:00 Last Friday, on day two of [AGNTCon + MCPCon North America](https://events.linuxfoundation.org/agntcon-mcpcon-north-america/), the [Agentic AI Foundation](https://aaif.io) gave me the **Outstanding Technical Contribution Award** for my work on the [MCP Python SDK](https://github.com/modelcontextprotocol/python-sdk). You can watch the moment on the day two livestream, right after the opening keynote. ## I'm one of the two maintainers of the MCP Python SDK. Python developers use it to build MCP servers and clients, and it's where a lot of my time has gone this year. The biggest piece of that work was v2. We shipped the [v2 beta](https://pydantic.dev/articles/mcp-python-sdk-v2-beta) alongside the 2026-07-28 revision of the MCP specification, with a new `MCPServer` API and multi-round-trip requests. v2 is now the stable release line. A lot of v2 is the new specification, implemented in Python: standard error codes, authorization hardening, routable transport headers, a deprecation policy, and OpenTelemetry tracing on by default. On the specification side, I sponsored [SEP-1288](https://github.com/modelcontextprotocol/modelcontextprotocol/issues/1288), the WebSocket transport. The HTTP transport also runs on [Starlette](https://github.com/Kludex/starlette) and [Uvicorn](https://github.com/Kludex/uvicorn), which I maintain, so when the SDK finds a bug underneath it, I can fix it in the layer where it belongs. Most of that work happens in issues, pull requests, and spec discussions that nobody sees on a conference stage. It's nice to have it noticed. ## First, a special thank you to my colleague [Laís Carvalho](https://pydantic.dev/authors/lais-carvalho). She nominated me, and I would not have been on that stage without her. To the MCP maintainers and contributors: the SDK is better because of every issue, review, and pull request you sent. To [Pydantic](https://pydantic.dev): you pay me to work on open source, and you let me spend that time where the community needs it. That is rarer than it should be. It also goes both ways. I work on [Pydantic Logfire](https://pydantic.dev/logfire) too, so v2 ships an OpenTelemetry middleware enabled by default. Call `logfire.configure(distributed_tracing=True)` in your MCP server and `logfire.configure()` in your client, and every request between them arrives in [Logfire](https://pydantic.dev/logfire) as one trace. The [v2 beta post](https://pydantic.dev/articles/mcp-python-sdk-v2-beta) walks through it. I also shared the stage with our friends Antoni Kozelski and Bartosz Gonczarek from [Vstorm](https://vstorm.co/?utm_source=pydantic&utm_medium=partnership&utm_campaign=aaif-outstanding-technical-contribution-award), who won the **Open Agentic AI Innovation Award**. Congratulations to both of you, and thank you for making the photo a lot better. The [Logfire MCP server](https://pydantic.dev/docs/logfire/guides/mcp-server/) is where this work lands first. It runs on the same Python SDK, so the standard error codes, the authorization hardening, and the tracing middleware above all show up there before anywhere else. I want it to be the best MCP server you connect your coding agent to, and I spend a lot of my week making that true. If you have a Logfire project, add `https://logfire-us.pydantic.dev/mcp` (or the EU URL) to your agent and ask it what broke in production last night. Setup takes a few minutes, and the answer usually takes less. See you at the next MCPCon.