Clawmetry

ClawMetry is an observability and governance layer for AI agents. It records what coding and autonomous agents are doing across files, commands, tool calls and model spend, flags risky patterns such as stuck loops, dangerous commands and spend spikes, and lets teams apply policies that pause a looping run, hold a risky tool call for approval or cap spend through a proxy. It installs with a single pip command and runs locally by default. It suits developers and teams running agents such as Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Aider, Devin and OpenClaw, with 32 supported runtimes in total and no SDK needed for supported local runtimes. Notifications go to Slack, PagerDuty, WhatsApp and other connected channels. Activity content is encrypted on the user's machine with client side AES encryption under a key the servers do not store, while operational metadata such as hostnames, model names, token totals and tool names syncs in readable form. It can run in ClawMetry Cloud or self hosted offline. Pricing is per node. Free is $0 forever for one node on the OpenClaw, NVIDIA NemoClaw, Goose and Qwen Code runtimes, with no account needed. Starter, the observability tier for all 32 runtimes, is $9 per node per month or $90 per node per year, shown against a $19 list price. Pro, which adds policy gates, evals and waste detection, is $19 per node per month or $190 per node per year, shown against $39. Both include a seven day free trial, and annual plans include the $149 ClawMetry desk device free. Enterprise is an annual contract with custom pricing, adding audit, SSO and a customer hosted server. The core package is MIT licensed, while the Pro extension is proprietary. ClawMetry is published by Instalabs LLC and is a founder led operation with no separate support team.