> ## 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. # Introduction > Claw Lens is a zero-config local dashboard for your OpenClaw AI agents — observability, cost analytics, security audit, and debugging tools in one place. Run `npx claw-lens-cli` to start the dashboard at `http://localhost:4242`. No account, no API key, no external service — everything runs on your machine. Available in both English and Simplified Chinese. Claw Lens Overview Dashboard *** ## You type a prompt. Your agent runs for five minutes. What is it doing? You see the final output — but not the tool calls, the retries, or the tokens it burned to get there. Right now, running an AI agent feels like flying blind. *** ## Live Monitor Watch your agents work in real time. See which agent is running, what tool it just called, whether it's stuck in a loop — streamed live over WebSocket, not polled from logs. Every message, every tool call, every result, as it happens. Live Monitor — real-time agent activity stream *** ## Cost & Token Analytics Track spend by day, by model, and by agent. See cache hit rates, cost-per-session breakdowns, and token composition across any time range. When a session costs more than expected, drill into the per-agent breakdown to find out where the tokens went. Token Usage — cost summary and trend Token Usage — per-agent and per-model breakdown *** ## Sessions Browse every session in one filterable table — model, tokens, cost, cache hit rate, context pressure, burn rate, errors, and activity time at a glance. Filter by agent, time range, or search by session ID. Sessions — filterable session table Click into any session to see the full details: cache hit analysis, cost and token charts that grow step by step, context pressure over time, and every user and assistant message with token in/out, cache read/write, cost, and latency. Expand any message to inspect arguments, output, and errors. Session Details — per-step cost and token breakdown Session Messages — message-level token and cost details *** ## Session Timeline A Gantt chart of a single session — every LLM call and tool invocation laid out on a time axis. See exactly where time is spent: is it the model thinking, or a tool running slow? Idle gaps are auto-compressed so you focus on what actually happened. Session Timeline — Gantt-style execution trace Click any turn to expand the full detail — input arguments, output, and errors. Session Timeline — Gantt-style execution trace *** ## Agents See all your agents at a glance — who's running, who's idle, who's healthy. Each agent card shows health status, current task, model, cost, error rate, cache hit rate, and context pressure. Agents — health, status, and activity overview Expand any agent to see token usage trends, a 7-day activity heatmap, session frequency, and error history. Agents — expanded detail with trends and heatmap *** ## Security Audit Automatic scanning across every session for credential leaks, prompt injection attempts, dangerous shell commands, and data exfiltration patterns. Each event is risk-scored so you can filter by severity. Includes a per-agent security profile — tool footprint, directory access, network domains, and a trust status. Security Audit — risk-scored event timeline *** ## Context Breakdown Visualize how system prompt, conversation history, and tool results fill up the context window at every API call. See exactly when and where context pressure spikes. Context Breakdown — context window composition per step *** ## And more Auto-detect scheduled agent runs with full run history — status, duration, and success/failure logs. Latency distribution, failure rates, and slow call identification for every tool your agents use. Detect turns with abnormally deep tool-call sequences — the first sign an agent is looping or struggling. Pipeline flow visualization and diff between consecutive LLM calls to debug prompt cache misses. Full dashboard available in English and Simplified Chinese. Switch anytime from the sidebar. A little easter egg — find the mascot hiding in settings and adopt it. It comes with different moods. *** ## Get started in 30 seconds ```bash theme={null} npx claw-lens-cli ``` The dashboard opens at `http://localhost:4242`. No config, no signup. Claw Lens auto-detects your OpenClaw home directory and reads agent logs locally. All data stays on your machine. Requires Node.js 18+. Your OpenClaw agent must have run at least one session to produce log data. *** Detailed setup guide and configuration options. Explore every feature in depth.