

Battle of the Agent Runtimes: Claude vs OpenAI (and the rest)
Everyone’s building agents. Fewer people are asking what they actually run on.
The runtime you pick decides how your agents handle tools, state, memory, long-running work and cost, and it decides how hard it’ll be to leave later. In this session we’ll put Claude Managed Agents and OpenAI’s agent runtime head to head on the same task, live, and dig into where each one shines and where it bites.
We’ll also cover:
Managed and opinionated runtimes versus frameworks that give you more control, like LangGraph, and when to choose which
The trends shaping the space, from MCP and multi-agent orchestration to evals and lock-in
Why your agents’ context shouldn’t belong to whichever runtime wins, and how we handle that at Coconut
It’s a bit more technical than our last session, but practical rather than academic. Bring your questions and your war stories.