Introduction to Mission Control Agent
When an inference endpoint slows down, finding the cause usually means working through dashboards, metrics, and deployment config by hand. Mission Control Agent is built into the CoreWeave Console. You describe the problem in plain language, the agent investigates your telemetry, and it can make the fix itself.
In this demo, a GLM-5.2 endpoint is returning slow responses. The agent traces the problem to KV cache pressure, switches the model to NVFP4 weights, then splits it into two tensor-parallel replicas. Throughput goes up on the same eight GPUs, with no new hardware.
