Traccia open-sources Agent runtime governance SDK

Traccia has released an open-source Python SDK for observing, evaluating, and enforcing policies on AI Agents and LLM applications at runtime. It provides distributed tracing and token and cost statistics based on OpenTelemetry, and uses the @govern decorator to deny or rewrite LLM and tool calls according to policies such as Spend Cap, Model Boundary, and Loop Cap. The SDK has been published on PyPI, but runtime policy enforcement for @govern requires the Traccia platform: users need to create a workspace API key in Settings → API Keys. Users who only need tracing or self-hosting should use @observe instead, which only creates trace spans and can be used with any OTLP backend. If the agent ID, API key, or endpoint is missing, @govern throws a GovernanceConfigError (a subclass of AgentBlockedError) rather than continuing to run without governance.

AI overview based on source reports · Updated 2 days ago