Hi Product Hunt! I'm Sara, the maker of claw-lens. I've been running OpenClaw agents for months and kept hitting the same two frustrations. First, I had no idea where my money went. I'd look at my API bill and see a number with no way to break it down. Which agent? Which model? Was it a cron job running every 5 minutes, or one session that went off the rails? I couldn't tell. Second, I had no idea what my agents were actually doing. A task that takes 5 minutes is 5 minutes of darkness. If I ask "how's it going?", the answer varies — sometimes it's stuck, sometimes it's done, sometimes it's still working. And even when it says it's still working, it can't tell me what step it's on. I needed to see the execution stream, not ask for a status label. I looked at existing tools — most are cloud-based dashboards that want your agent data uploaded. For me that's a non-starter. My agents read local files, touch credentials, and execute shell commands. That telemetry doesn't belong on someone else's server. So I built claw-lens. It reads the JSONL files OpenClaw already writes to disk, parses them into SQLite, and serves everything locally. Zero config, zero outbound calls, zero accounts. Every page answers a question I actually had: Where did my money go? → Overview for the big picture; Tokens for the breakdown — by agent, model, and time (hour / day / week) across 7- or 30-day windows What is my agent doing right now? → Live Monitor — real-time view of every tool call as it happens Why did this session burn 800K tokens? → Click into the session for per-step cost, tokens, and raw user / assistant messages; Timeline lays every turn, tool call, and step on a Gantt chart How are all my agents doing? → Agents page — per-agent status, error rate, cost, and activity at a glance Did my agent access something it shouldn't? → Audit — risk-scored tool call timeline One command: npx claw-lens-cli Open source, MIT. Feedback and issues welcome: github.com/msfirebird/claw-lens

Claw Lens

Analyze, debug & audit your OpenClaw agents — fully local

Analyze, debug & audit your OpenClaw agents — fully local