Skip to main content
Log in Get started

Bring framework-neutral AI agent traces into the same evaluation loop.

If your agent, app server, LLM SDK, or workflow tool emits OpenTelemetry, export OTLP/HTTP JSON to Mibo and evaluate real interactions against the tests you already use before release.

Get started freeRead OTLP setup

OTLP/HTTP JSON is the handoff

Configure your exporter to send traces to Mibo's trace endpoint. Mibo receives the OTLP envelope, stores the trace, and evaluates active test cases against it as passive testing.

OTEL_EXPORTER_OTLP_TRACES_PROTOCOL=http/json
OTEL_EXPORTER_OTLP_TRACES_ENDPOINT=https://api.mibo-ai.com/public/traces
OTEL_EXPORTER_OTLP_TRACES_HEADERS=x-api-key=<your_api_key>
OTEL_SERVICE_NAME=<otlp_service_name>

Set the service name to match the OTLP service name on the Mibo agent. Mibo accepts OTLP/HTTP JSON, not protobuf or gRPC.

Observability becomes evaluation when the trace has expectations

Instrument the path

Capture the spans that explain the interaction: request, model response, routing decision, tool call, and final answer.

Keep tests framework-neutral

The ingestion path does not change the test suite. Your expectations stay useful if the underlying agent implementation changes.

Investigate evidence

Review semantic and procedural results with the trace context, then run the same scenario actively while you verify a fix.

Keep exploring