Your OpenClaw agents spawn sub-agents, burn tokens, call tools. You have no idea what they're actually doing. ClawMetry shows you, in real-time.
pip install clawmetry && clawmetry connect
macOS, Linux, Windows. The installer figures it out.
One screen. Everything happening across your agents.
Built because we needed it. Now you don't have to build it yourself.
Your agent says "spawning sub-agent" and you just... hope it works? ClawMetry shows you every step: what files it's reading, what commands it's running, what tools it's calling, and what it's thinking. Summary, narrative, full logs. No more guessing.
Sub-agent running
Sub-agent on completion
Channels, gateway, models, tools, nodes. All wired up, updating as things happen.
Tokens in, tokens out, cache hits, response times, cost per call. Numbers you actually need.
Cron jobs, service uptime, disk usage, active sub-agents. Glanceable.
Every session logged. Timeline, tool calls, cost. Go back and see what happened.
Hooks into OpenClaw's task system. See which agents are on which tasks.
Per-session, per-model, per-tool. Know what you're spending before the invoice shows up.
If OpenClaw runs there, ClawMetry runs there. Laptop, server, Raspberry Pi, whatever.
Skip the setup. We host it, you just watch.
Get notified about new features, releases, and the growing -metry ecosystem.
Everything you need to know about ClawMetry.
ClawMetry is a free, open-source real-time observability dashboard for AI agents running on OpenClaw. It gives you instant visibility into token costs, cron jobs, sub-agent activity, memory file changes, and session history — all in a single local dashboard.
Yes, completely. ClawMetry is MIT licensed and runs locally on your machine. No cloud dependency, no telemetry, no subscription. There is also an optional ClawMetry Cloud plan for teams who want a hosted multi-node dashboard.
Run pip install clawmetry then clawmetry. It auto-detects your OpenClaw workspace — no config file needed. Setup takes about 30 seconds.
ClawMetry monitors token usage and cost per session and model, cron job status and history, sub-agent spawns and outcomes, memory file changes (SOUL.md, MEMORY.md, AGENTS.md), session transcripts, channel activity, and real-time tool call traces.
Yes. ClawMetry works with any model your OpenClaw agents use — Claude (Anthropic), GPT-4 (OpenAI), Gemini (Google), and others. It tracks cost and token usage per model automatically.
LangSmith and W&B are enterprise platforms built for LLM application developers. ClawMetry is purpose-built for OpenClaw users — it understands the specific concepts of channels, cron jobs, sub-agents, memory files, and the full agent runtime. It runs locally in 30 seconds with zero config.
ClawMetry runs on Linux, macOS, and Windows (via WSL). It also works on Raspberry Pi and other ARM devices. Requires Python 3.8+ and an active OpenClaw installation.
