Observability

One trace across every agent hop.

Every call is instrumented automatically (LLM, MCP, agent-to-agent, or internal API) and pushed to the observability stack you already use.

Fabric instruments each hop and emits standard OpenTelemetry, so traces land in the stack you already run.

1: trace ID across every hop
0: application-code changes to instrument
OTEL: standard format, no new dashboard

The problem

Agent workflows fan out across models, tools, and services. Reconstructing what actually happened after an incident means stitching together logs from five different systems that don't share a trace ID.

How it works

Fabric instruments every hop as it happens, tagging traces so a single agent action can be followed end to end, then emits that data in standard OpenTelemetry (OTEL) format. Because it's OTEL, it pushes straight into whichever stack you already run (Datadog, Grafana, Honeycomb, or your own collector) with no proprietary dashboard to adopt in between.

What you get

Automatic instrumentation

No application-code changes required.

Standard OTEL output

Pushes to your existing observability stack instead of a new one.

Single trace ID

One ID spans every hop in a multi-agent workflow.

Every workload type

APIs, MCP servers, and LLM calls, not just one.

Token-aware metrics

Combine with token metrics for per-team cost attribution.

Bring every AI interaction under control

Connect your first model, MCP server or internal API and apply consistent governance from one gateway.