Monitor agents

agenttop

Like htop, but for agents: every running agent with its resource use, model usage and the scheduling policy currently in use.

agenttop

List agents

agentctl ps                 # agents and their instances
agentctl members <agent>    # the processes that make up one agent
ls /agent/local             # the same agents, as a directory

Execution traces

Every task leaves a full execution trace: model calls, tool calls and system actions, step by step, with their time and token use.

agentctl trace

A trace shows, in order:

Step	Recorded
Model call	Model, tokens in and out, time
Tool call	Which agent and skill was called, with what result
System action	What changed on the system, and whether you were asked first

Use it to see exactly which step went wrong, or where the tokens went.

Stop an agent

agentctl kill <agent>

See the CLI reference for every command.