ClawMetry

Real-time observability and governance layer for AI agents across 33 runtimes.

Overview

ClawMetry is a control plane for AI agents that provides real-time observability and governance across 33 supported agent runtimes including Claude Code, Cursor, GitHub Copilot, OpenClaw, and many others. It reads session data that agents already write to disk, stores it locally in DuckDB, and surfaces activity, spend, and risk signals through a unified dashboard.

The platform helps teams discover what agents are running on their machines, observe what they're doing and how much they cost, and govern their actions through optional controls like pre-tool approval gates, budget limits, and process controls. ClawMetry is local-first by default with no SDK required—it works by reading existing agent session stores without modifying how agents run.

Key capabilities include detecting wasteful patterns like stuck loops, tracking token spend and costs, approving or blocking tool calls before execution, and maintaining an audit trail of all decisions. Controls are opt-in and off by default, giving teams fine-grained control over when and how ClawMetry intervenes.

Key features

Support for 33 agent runtimes
Local-first architecture with no SDK required
Real-time activity monitoring and session replay
Token and cost attribution per agent
Stuck loop and waste detection
Pre-tool approval gates
Budget enforcement via optional proxy
Process-level pause and stop controls
Audit trail and tamper-evident records
MCP server integration for agent self-diagnosis

Pros

Works with existing agents without code changes
Local data storage by default
Open-source core with 434 GitHub stars
Covers diverse agent ecosystem (33 runtimes)
Fine-grained control over when enforcement acts
Detailed cost and token attribution
No vendor lock-in for observation

Cons

Enforcement capabilities vary by runtime
Per-machine installation required for fleet rollout
Some controls require specific runtime support
Pre-tool gating only available on certain runtimes
Mid-run process control not available on all platforms

Use this if

You need to monitor and control multiple AI agents across your organization, want to understand agent spend and detect wasteful patterns, or need governance controls like approval gates and budget limits.

Skip this if

You only use a single agent in isolation and don't need observability or governance, or your agent runtime is not among the 33 supported platforms.

Best for

Teams deploying AI agents across their organizationMonitoring agent spend and token usageDetecting and preventing runaway agent behaviorAuditing agent activity and decisionsGoverning tool access and model callsInvestigating failed or inefficient agent sessions

Alternatives

Langsmith (LangChain observability)Arize (ML observability)Datadog (general infrastructure monitoring)Custom logging solutions

Compare ClawMetry alternatives

More AI

Compare all

coolplugz

Orchestration layer that guides Claude Code to deliver coding tasks without constant supervision.

Goldfish

AI assistant that remembers your work and writes replies in your tone, available for macOS and Windows.

Runtime

AI agent platform for payment teams to automate operations, risk, compliance, and support workflows in their own cloud.

Scribble AI

AI-powered blog writing tool that generates SEO keywords, outlines, first drafts, and hero images from a title.