What's new in AgentMetrics

Added

• Python SDK with zero-config agent run tracking via @agentmetrics.track() decorator
• Automatic token usage capture for OpenAI, Anthropic, and Google Gemini models
• Per-run cost calculation based on current model pricing, updated weekly
• Tool call tracking: captures tool name, input, output, latency, and success/failure per call
• Subagent tracing: parent-child run relationships are preserved and visualized in the dashboard
• Real-time dashboard with run timeline, cost trends, error rate, and p95 latency charts
• OpenClaw plugin: install with a single CLI command, zero code changes required
• Hermes integration: wrap any Hermes agent with @agentmetrics.track() to capture full run telemetry
• REST API for querying runs, aggregating cost by time range, and exporting raw events
• API key management: create, rotate, and revoke keys from the Settings page
• Webhook delivery for run.completed and run.failed events with HMAC-SHA256 signatures
• 7-day run retention on the Free plan, 365 days on Pro