---
id: clawmetry
version: "0.12.704"
license: MIT
license_treatment: permissive
maintenance: active
---

# clawmetry — ClawMetry - Real-time observability for 12 AI agent runtimes (OpenClaw, NVIDIA NemoClaw, Claude Code, Codex & more)

License: permissive · Maintenance: active · Downloads: 93.1K/mo

## What it is and what it does

ClawMetry is a monitoring dashboard that connects to AI agent runtimes running on your machine and displays their activity in real time. It auto-detects supported runtimes (OpenClaw, NVIDIA NemoClaw, Claude Code, Codex, Cursor, Aider, Goose, and others) and aggregates their session data into a single web interface served on localhost:8900. The dashboard shows live message flow diagrams, token usage and cost breakdowns, session histories, scheduled jobs, real-time logs, workspace memory browsing, and alert/approval controls.

The package includes a Flask-based backend with DuckDB for local data storage and WebSocket support for live updates. It can also install hooks into Claude Code to gate tool calls behind manual approval before execution. Most runtimes beyond OpenClaw and NemoClaw are available through a paid license tier, though the free version covers the primary open-source runtimes. Installation is a single command with zero configuration needed.

Use it for:

- Monitor token usage and costs across multiple AI agent sessions in one dashboard
- Set budget caps and error-rate alerts that route to Slack, Discord, PagerDuty, or email
- Gate risky tool calls in Claude Code behind manual approval with phone push notifications
- Browse agent session transcripts, memory files, and audit logs for debugging or compliance
- Track active agent sessions and scheduled jobs across your entire agent fleet

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

ClawMetry provides a real-time observability dashboard for AI agent runtimes, auto-detecting and monitoring 14 different agent systems (OpenClaw, NVIDIA NemoClaw, Claude Code, and others) from a single unified interface.

Yes, if you run AI agents (especially OpenClaw or NVIDIA NemoClaw) and want centralized observability. Install friction is low, maintenance is active, and the MIT license imposes no restrictions. Most other runtimes require a paid license, so verify your agent's support tier first. No known security vulnerabilities.

## Install

pip install clawmetry
uv add clawmetry
poetry add clawmetry

## Installing clawmetry

Before you install: Low friction install with a single pip command; the package is actively maintained with a recent release and depends on common web and database libraries (flask, waitress, duckdb, websocket-client, cryptography, truststore).

License in practice: MIT license permits commercial and private use with minimal restrictions; you may use, modify, and distribute ClawMetry freely as long as you retain the license notice.

Quickstart:

pip install clawmetry
clawmetry
# Opens dashboard at http://localhost:8900

Requires Python 3.8 or later; agent runtimes must be installed and running on the same machine for auto-detection to work.

Verify before relying:

- Whether the 14 agent runtimes listed are all fully supported in the free tier or if most require a paid license
- Whether the cloud features (push notifications, cloud snapshots) require external account setup
- Performance characteristics when monitoring multiple concurrent agent sessions

## Package facts

- License: MIT (permissive)
- Python support: supports_current
- Install friction: low
- Maintenance: active
- Downloads: 93.1K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

ai agent observability dashboard, real-time agent monitoring, openclaw observability, agent runtime tracking, ai agent fleet monitoring, agent session analytics, ai tool call tracing, ai-observability, agent-monitoring, dashboard

[View on SkillFed](https://skillfed.io/packages/clawmetry) · [View on PyPI](https://pypi.org/project/clawmetry/)