Skip to main content
OpenClaw is a framework for building and running AI agents. Claw Lens is an observability  and devtool built specifically for OpenClaw — it reads the session files that your agents write and surfaces them in a local dashboard.Without OpenClaw, there are no agent sessions for Claw Lens to read, so the dashboard will not have any data to show.
No. Claw Lens reads session files produced by OpenClaw agents. If you have not run any OpenClaw agents, there are no session files to read and the dashboard will show an empty state.If you are just getting started, see the OpenClaw documentation first, then come back to Claw Lens once your agents have produced some sessions.
All data is stored in a local SQLite database at ~/.openclaw/claw-lens.db (or $OPENCLAW_HOME/claw-lens.db if you have set that environment variable). Nothing is sent to any external service.The database holds sessions, messages, tool calls, security audit findings, and agent risk scores. See the Data Storage page for full details.
If you use npx: you already get the latest published version every time you run npx claw-lens-cli. No action needed.If you installed globally: run the following to upgrade to the latest version:
npm install -g claw-lens-cli@latest
There are a few common causes:
  1. Your agents haven’t run yet. Claw Lens can only show sessions that have already been recorded. Run an OpenClaw agent first, then reload the dashboard.
  2. OPENCLAW_HOME is set incorrectly. If you have set this environment variable, make sure it points to the directory that actually contains your agents/ folder. Claw Lens looks for session files at $OPENCLAW_HOME/agents/*/sessions/*.jsonl.
  3. The data hasn’t been refreshed. Try clicking Refresh Data on the Settings page, or send a POST request to http://localhost:4242/api/refresh. This forces Claw Lens to re-read all session files.
  4. Session files are empty or malformed. If an agent was interrupted before it produced any messages, Claw Lens has nothing to show for that session.
In the Live Monitor, an agent is marked as stuck when it has been running for a notable period without making any observable progress — meaning no new tool calls or LLM messages have appeared for an extended time.A stuck agent may be waiting on a long-running tool, blocked on I/O, or in an unresponsive state. The Live Monitor surfaces this so you can decide whether to investigate or interrupt the agent.
Claw Lens is designed for local development use — its primary purpose is giving you an interactive dashboard while you build and test agents. That said, you can run it in a non-interactive environment with:
claw-lens --no-open --port 4242
Use --no-open to skip the browser launch, and set a custom port if needed. Keep in mind that the dashboard binds only to 127.0.0.1, so it will not be reachable from outside the machine even in a CI context.
Delete the SQLite database file and restart Claw Lens:
rm ~/.openclaw/claw-lens.db
npx claw-lens-cli
If you use a custom OPENCLAW_HOME:
rm "$OPENCLAW_HOME/claw-lens.db"
npx claw-lens-cli
On the next startup, Claw Lens creates a fresh database and re-reads all your agent session files from scratch.
This also removes all security audit findings, dismissed false positives, and computed risk scores permanently.
Claw Lens requires Node.js 18 or later. Run node --version to check what you have installed. If you need to upgrade, visit nodejs.org or use a version manager like nvm.
No. The security audit uses pattern matching and heuristics to flag potentially sensitive content — things like credentials in prompts, risky shell commands, or prompt injection attempts. It will produce false positives, and it may miss things that a manual review would catch. This is on par with similar tools and skills in the space — no automated system catches everything, but it surfaces the issues you’re most likely to miss on your own. We are actively improving detection coverage and accuracy.Treat the audit as a starting point for investigation, not a definitive verdict. Review each flagged item manually and use the Dismiss action to mark false positives so they stop appearing in future reports.