1 comment

[ 1028 ms ] story [ 385 ms ] thread
I’m curious how you're handling rate-limit backpressure at the MCP layer when an LLM blasts multiple tool calls at once.