After initialization you can start capturing traces. Quotient provides two primary approaches:
  1. @quotient.trace decorator – wrap your entrypoint or key functions to record spans.
  2. OpenInference instrumentors – automatically instrument supported frameworks and providers.

​
Decorate an Entrypoint

traced_agent.py
from quotientai import QuotientAI

quotient = QuotientAI(api_key="your-quotient-api-key")

quotient.tracer.init(
    app_name="my-app",
    environment="dev",
)

@quotient.trace('support-agent')
def run_agent(question: str) -> str:
    """Main entrypoint for your agent."""
    # call tools, models, etc.
    return "Answer goes here"

run_agent("How do I reset my password?")
The decorator creates a root span for the function and tracks child spans for tool calls when instrumentation is available.

​
Use OpenInference Instrumentation

Quotient integrates with OpenInference so you can reuse the same telemetry across frameworks. Attach the appropriate instrumentor when initializing the tracer:
with_instrumentor.py
from openinference.instrumentation.langchain import LangChainInstrumentor
from quotientai import QuotientAI

quotient = QuotientAI(api_key="your-quotient-api-key")

quotient.tracer.init(
    app_name="langgraph-weather-app",
    environment="dev",
    instruments=[LangChainInstrumentor()],
)
Instrumentors add spans for prompts, tool invocations, retriever calls, and more—no code changes required in the rest of your app.

​
Supported Frameworks

We provide step-by-step guides for popular agent frameworks: Looking for another framework? See OpenInference instrumentation for additional options.

​
Tracing Without OpenInference

You can still use the decorator without OpenInference instrumentors. You will capture the function-level span and any custom spans you emit manually. Note that certain Quotient features (detections, tooling insights) rely on the richer span data provided by instrumentors.
Next: Custom Spans