1 comment

[ 3.6 ms ] story [ 24.5 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.