Same here. I just refreshed the page and retyped my prompt and it worked. It was a weird occurrence though. I’m assuming it’s some artifact of a back end error.
I’ve seen similar behavior with extended-output models where the generation appears to get stuck repeating the same status instead of completing. It may be related to a generation or streaming issue rather than the prompt itself.
If it happens repeatedly, it would be useful to note the prompt, model settings, and whether it eventually stops after a timeout.
This is an important topic as agentic systems become more common. Traditional API gateways can handle standard requests well, but MCP introduces another layer where tool calls and payloads need to be governed dynamically.
The ParsePayload policy and payload operation controls sound especially useful for getting better visibility and control over which tools agents can actually invoke. Looking forward to seeing the architectural patterns discussed in the session.