Logs
ExoHuman records what agents do at three levels: executions (every run, with its transcript), container logs (the agent’s process output) and resource metrics.
Executions
Section titled “Executions”Every chat message, task, scheduled run, loop iteration and agent-to-agent call is an execution. The agent’s Tasks tab lists them with trigger, status, duration, model, cost and context usage; open one to see the full transcript — messages, tool calls and results — including live progress while it runs.
Fleet-wide, admins see executions across all agents with filters by agent, status and trigger, plus a timeline view (also available from GET /api/executions and GET /api/executions/timeline).
Container logs
Section titled “Container logs”Container logs are the raw stdout/stderr of the agent’s container — startup, runtime messages, errors. They are useful when an agent fails to start or a tool misbehaves. Owners and admins can read them through:
| Surface | How |
|---|---|
| REST API | GET /api/agents/{name}/logs?tail=100 |
| MCP | get_agent_logs(agent_name, lines) |
| CLI | exohuman logs <agent> |
Guardrail denials and credential-leak detections are written to the agent’s own guardrail log and the audit log.
Resource usage
Section titled “Resource usage”The agent header shows live CPU, memory and uptime, refreshed every 10 seconds while the agent is running. The same data is available from GET /api/agents/{name}/stats.
On ExoHuman Cloud, the minutes an agent runs on its compute profile are what you are billed for; see usage per agent and profile in Billing.
Each execution records its model cost (estimated from token usage reported by the runtime). The Overview tab rolls this up per agent, and admins can see fleet-wide cost and model spend per team. Model spend through the model gateway can be capped with team budgets.
Retention
Section titled “Retention”Execution history and logs are kept according to your instance’s retention settings; the audit log is retained for 365 days. Export what you need to keep longer.