Archestra-Grafana integration

Archestra integrates with Grafana to provide comprehensive observability for AI agent deployments. The platform exports metrics via Prometheus and includes pre-configured dashboards for monitoring.

How it works

Archestra exposes Prometheus metrics on /metrics endpoint including:

• llm_tokens_total — token usage across all LLM providers
• llm_request_duration_seconds — LLM API call latency
• http_request_duration_seconds — HTTP API performance
• Tool blocking events and security guardrail triggers

Use cases

• Monitor AI spending across teams and agents
• Track security events and blocked tool calls
• Visualize LLM performance metrics (time to first token, tokens per second)
• Set up alerts for anomalous agent behavior

Configuration

1. Configure Prometheus to scrape Archestra's metrics endpoint
2. Import Archestra's pre-configured Grafana dashboards
3. Set up alerts for budget limits or security events

Tradeoffs

• Requires additional infrastructure (Prometheus + Grafana) for full observability
• Metrics add minimal overhead (~45ms gateway latency)