AI Agents
AI Agent Friendly is one of three core design pillars of Claw Insights, alongside Zero Intrusion and Shareable Snapshots.
Why does an AI agent need observability? Because agents that can see their own metrics make better decisions — they know when token usage is spiking, when sessions are failing, and when it's time to generate a status report.
How Agents Interact with Claw Insights
1. Agent Skills — Drop-in Discovery
Skills are structured instruction files that agents automatically discover and use. Drop them into your agent's skill directory — the agent learns how to install Claw Insights, generate snapshots, and query status without any prompt engineering.
Best for: agents that support skill discovery (like OpenClaw).
2. Programmatic Access — Direct API Calls
Use the GraphQL API to query sessions, metrics, and system health. Use the REST API to generate visual status cards. All endpoints are curl-friendly.
Best for: custom integrations, scripts, and agents that call APIs directly.
→ See the Programmatic Access guide for end-to-end examples.
Common Scenarios
| What you want | How to do it |
|---|---|
| "How's my token usage?" | Query source.metrics via GraphQL |
| Daily status card to Slack | Cron + POST /api/snapshot → send to channel |
| Alert on error spikes | SSE subscription on dataChanged signal |
| Review sub-agent activity | Query source.sessions + sessionTranscript |
| Quick dashboard capture | claw-insights snapshot -o status.png via CLI |