The Overview page is the home screen of Claw Lens. It pulls live data from your OpenClaw agents and presents the most important signals without requiring you to navigate between pages. From here you can spot cost spikes, failing agents, and stuck sessions at a glance.Documentation Index
Fetch the complete documentation index at: https://claw-lens.com/llms.txt
Use this file to discover all available pages before exploring further.
KPI strip
The top row shows five key metrics for today.
| Metric | What it measures |
|---|---|
| Cost today | Total LLM spend since midnight, with a 7-day total and a week-over-week change indicator |
| Tokens today | Total tokens consumed since midnight, with a 7-day total |
| Sessions today | Number of sessions started since midnight, with a 7-day count and message total |
| Sessions with errors today | Sessions that contain at least one error message; shown in red when non-zero |
| Cache efficiency | Proportion of input-side tokens served from the prompt cache (cache_read / (cache_read + input_tokens)) |
Agents panel
The Agents panel lists every active OpenClaw agent. Agents are sorted with errors first, then warnings, then healthy — so the agents that need attention appear at the top.
- A status dot colored by health verdict (green / amber / red)
- The agent name and its health badge (
healthy,warning, orerror) - The primary model (shortened for readability, e.g.
sonnet-4-5instead ofclaude-sonnet-4-5-latest) - Today’s cost for that agent
- A context bar showing how full the context window was in the last session (green → amber → red as it approaches the limit)
- The last activity time and current operational status
Health Status
| Status | Meaning |
|---|---|
healthy | No recent errors in the last 10 messages of the most recent session, and context usage below 80% |
warning | 1–2 tool failures, or the agent hit max_tokens in the last 10 messages; or context usage 80–99% |
error | 3+ tool failures, a session that ended with stop_reason: error, or context usage ≥ 100 % |
Operational states
| State | Color | Meaning |
|---|---|---|
running | Green | The agent’s session file was modified within the last 30 minutes |
stuck | Amber | The session file was recently modified but the agent has recent tool failures or an error stop reason |
idle | Blue | No session file activity in the last 30 minutes |
stale | Gray | The agent directory exists but there are no recorded sessions |
Models panel
The Models panel shows which LLM models are currently in use across all agents.
- An active/idle badge: a model is considered active if any agent using it had activity within the last hour
- The agent names that use this model as their primary model
- Today’s cost and 7-day cost for that model
Recent sessions
The Recent Sessions panel lists the five most recently active non-cron sessions, sorted by last message time.
- A status dot (green = no errors, red = has errors)
- Agent name, token count, and cost
- An optional task summary snippet — a short description of what the session was working on, extracted from the session file and cleaned of system prefixes
Cron jobs
The Cron Jobs panel gives you a quick overview of your scheduled agents.
- How many cron jobs are enabled
- How many are currently failing (consecutive errors > 0)
- When the next scheduled run is
- A status dot (green = last run OK, red = consecutive failures)
- The job name and its cron expression (e.g.
0 9 * * 1-5) - Duration of the last run
- When it last ran
Live Monitor mini-panel
- A pulsing green dot (running) or amber dot (stuck)
- Agent name, session ID, and status badge
- How long ago the session last had activity
- The current step trace: the last user message, any tool calls in progress, and the most recent AI response
Activity timeline
Timeline data is bucketed by calendar day in your local timezone. A day with no activity does not appear as a bar; hover over any bar to see exact values.
Data retention note
