An agent run isa trace, nota log file.
Agent frameworks that emit OpenTelemetry spans land here without a plugin. Every step becomes a span, every model call carries its token counts and finish reason, and a run that stalls names the tool call it stalled on.
Tool time and inference time are separate spans, so a slow run points at the right owner.
Included on every plan, including the free tier.
One span per step across planning, retrieval, tool calls and the final response.
Token counts and finish reason on every model call, split by model and operation.
Tool call latency separated from inference latency, so a slow run points at the right owner.
The same store as the rest of your telemetry, so an agent's database call is the database span your platform team already watches.
Three steps, then it is reporting.
What that looks like as a query.
{ name = "agent.run" && duration > 10s }
Runs that took too long, whatever the cause.
{ span.gen_ai.response.finish_reason = "length" }
Calls truncated by the context window.
sum by (gen_ai_request_model) (rate(gen_ai_client_token_usage_sum[1h]))
Which model the token spend is going to.
What AI teams ask before the first run.
Something not here? Book thirty minutes with an engineer who works on the platform. Book a call →
- Which agent frameworks do you support?
- Any framework that exports OpenTelemetry spans, with no plugin to wait for. Point its OTLP exporter at the agent on 4317 and the run arrives as a trace. Where a provider SDK has no instrumentation package yet, wrap the call in a span of your own.
- Do you store our prompts?
- Only if you switch it on. Message content is off by default and needs an explicit GENAI_CAPTURE_MESSAGE_CONTENT to enable. Prompts carry personal data, credentials people paste and whatever retrieval fetched, so turn it on one service at a time or leave it off.
- Can we see what a run costs?
- Token counts arrive split by model, by operation and by input against output. The rate stays yours, in the query or in a recording rule, because prices differ by provider, tier and contract.
- Does our application need an xScaler token?
- No. The agent holds the credentials. Your application knows one thing, the agent's OTLP endpoint on 4317 or 4318, and the agent adds the auth and tenant headers on the way out.
Start on the free tier.
20K active series, 50 GB of logs and 50 GB of traces a month, 90-day retention, and the whole of Insights.