Stop guessing why your agents failed. Restore memory state to the exact millisecond of failure — inspect variables, check objects, and debug like you're in a Jupyter notebook at the moment of crash. Agent tool execution is a blackbox. LLMops / tracing only logs inputs and output, not state. Code runs in secure VMs — Your functions execute in isolated Firecracker microVMs with automatic checkpointing Add @tool decorator — Mark functions that should run in BlickState's secure environment