Agent runs and usage

Every execution of an agent over an e-mail, with its tokens, cost and duration — plus two aggregate views over the same data. All read-only. See the AgentRun model and the Usage model.

Command	HTTP	Description
agent-runs list	GET /agent-runs	List agent executions with tokens, cost and duration.
agent-runs get	GET /agent-runs/{id}	Fetch one run, including the prompt version and any error.
usage summary	GET /usage/summary	Aggregated usage for a date range. Defaults to the last 30 days.
usage cost	GET /usage/cost	Cost broken down by model for a date range.

Examples

# List runs of one agent
kobana inbox agent-runs list \
  --params '{"agent_id": "AGENT_ID", "status": "failed"}' \
  --fields "id,status,total_cost,duration_ms"

# Get one run
kobana inbox agent-runs get \
  --params '{"id": "AGENT_RUN_ID"}' \
  --fields "id,status,error,input_tokens,output_tokens"

# Usage summary for a date range
kobana inbox usage summary \
  --params '{"from": "2026-08-01", "to": "2026-08-31"}'

# Cost by model
kobana inbox usage cost \
  --params '{"from": "2026-08-01", "to": "2026-08-31"}'

# Stream every failed run with automatic pagination
kobana inbox agent-runs list \
  --params '{"status": "failed"}' \
  --page-all --fields "id,agent_id,email_id,error"

Next steps

• Agents — the agent that produced these runs.
• Workflow runs — the HTTP chain a run's workflow triggered.