Product · heronsentry

HeronSentry

HeronSentry is a standalone observability plane for AI Agents: capturing runtime events via standard OTLP to deliver call chain traces, cost aggregation, alerts, and performance exports without depending on other suite products.

HeronSentry

Observability PlatformStandalone AI Agent observability plane: tracing, costs, alerts, and performance analysis

What Agents do, how well they perform, and what they cost—visible at a glance via standard OTLP.

OTLP: Open protocol; semantic fields mapped via profiles
Full Ingestion: Unsampled by default
Synchronous Ingest: Persisted within request lifecycle
Standalone: Deployable with zero suite dependencies

Typical pains

Troubleshooting broken Agent workflows relies on raw logs without distributed call chains

Costs remain invisible until month-end invoices arrive after budgets are blown

No visibility into whether specific Agents are healthy or experiencing abuse

Capabilities

Standalone DeploymentZero suite dependencies. Built-in ingestion, PostgreSQL storage, alert engine, and debug view

Standard OTLP IngestionReceives OTLP traces / metrics / logs (HTTP and gRPC). Built-in profiles for OpenClaw, OpenLLMetry, Claude Code, Codex, Gemini CLI, and NodalOS. Other languages can report Spans via HTTP Push

Distributed Call Chain TracingReconstructs call chains by parent-child relationships and links within a single OTel Trace, rendered as Mermaid diagrams

Cost AggregationSplits billing by cache read and cache write while displaying reasoning tokens. Reasoning token consumption is displayed in real time without extra surcharge items

Agent Reliability MetricsStreaming time-to-first-token (TTFT) and truncation rate (outputs hitting length limits). Tool rejection rates and conversation turn counts are available when reported upstream

AlertsError rate, cost spike, ingestion backlog, and high-frequency tool usage can trigger outbound webhooks (URL required). Latency spikes, orphaned Spans, and inactive Agents log to alerts/debug views without firing webhooks by default

Domain-Enhanced AlertsOptional and off by default. When enabled, detects in-progress tasks that are stuck and consecutive review rejections. Not blocking-chain, cross-group stall, or budget-burn alerts

Policy Evaluator Availability AlertsDistinguishes an unreachable NodalOS policy evaluator from a local policy fallback; these are not the same class of event as a security intercept

OTLP Ingest Rate LimitOTLP HTTP and gRPC can share a rate-limit bucket (per API key or peer; off by default). This does not cover native ingest

Topology VisualizationVisualizes individual call chains as Mermaid topologies

Performance & Cost AnalyticsCosts aggregate by Agent or Program. Reliability metrics are provided in Agent rollups

Does / does not

Does

Standard OTLP ingestion; supported frameworks mapped via profiles
Agent / tool / model call chain within a single Trace
Cost breakdown by cache read / write with reasoning token visibility
Alerts with optional webhooks
Optional domain-enhanced alerts (off by default)
Agent reliability and cost / performance exports

Does not

Does not store business logs
Does not replace infrastructure monitoring
Does not embed approval workflows
Does not reuse Grafana / Prometheus / Jaeger backends

Integrations

NodalOS OTLP push (agentd requires own_otlp enabled)AULO alerts view (outbox polling)PathPilot cost writeback and runtime health (error rate / P95 / timeouts)OwlAudit export ingestion for traces, costs, and alerts

Related products

NodalOSHeterogeneous multi-Agent hosting platform and governance substrate (OS layer)OwlAuditAI Agent compliance audit and human-in-the-loop (HITL) governance AULOOne workspace for human–Agent work: inbox, conversations, council, and governance actions PathPilotAI Agent task execution control plane

Next steps

See the full product loop Compare alternatives Integration path Book a demo