SDK

SDK

Instrument your agents with the Traccia SDK — decorators, auto-instrumentation, exporters, metrics, cost tracking, prompts, and evaluate().

SDK Quickstart

Get started with Traccia SDK in under 5 minutes.

Installation

Install the Traccia SDK for Python or TypeScript (Node.js).

@observe() Decorator

Instrument your functions with the @observe() decorator.

Agent identity

How agent and environment names appear when sending traces to the Traccia platform.

Auto-Instrumentation

Automatically trace LLM calls and popular libraries.

Exporters

Configure where your traces are exported.

Metrics

Understand the metrics emitted by the Traccia Python SDK and how to use them in Prometheus and Grafana.

Pricing & Costs

How the SDK calculates LLM costs locally, how to keep your pricing snapshot fresh, and how SDK costs relate to platform-recomputed costs.

Guardrail Detection

Passively detect, classify, and audit guardrails in your agent traces — zero runtime enforcement, no changes to existing code.

Governance in the SDK

Observe vs govern, trace evidence, and runtime policy enforcement.

Prompts in the SDK

Load versioned prompts at runtime with cache, fallback, and span identity.

Evaluate in the SDK

Run a task and scorers over a dataset from Python or TypeScript. Save an experiment by default, or keep the loop local.

Advanced SDK Topics

Manual spans, sampling, rate limiting, and advanced configuration.

© 2026 Traccia.