Show HN: Traccia - Observability, Runtime Control & Audit for agents

Traccia, a Bengaluru-based startup, launched its AI agent observability and governance platform this week, announcing a three-month free trial for early users, per a Show HN post by author vijaypoudel on August 21, 2026.

The product ships in two layers. The first is an open-source SDK that plugs into existing observability stacks — including Grafana, Tempo, and Jaeger — requiring only a few lines of code to instrument. The second is a hosted platform that adds evaluation, runtime policy enforcement, and audit trails on top of standard tracing. Both are described as cloud vendor-agnostic and framework-agnostic, according to the Traccia website.

The team's stated premise, explained in the HN submission, is that existing LLM tracing tools stop at observability, which is sufficient for passive monitoring but falls short once agents begin taking autonomous actions. Traccia's runtime controls allow teams to define policies that govern what an agent can do during execution — not just log what it did afterward. Audit logging then provides a record of those actions for compliance or review purposes.

No funding details, team size, or founding year were disclosed in the available sources. The company identifies itself only as an Indian startup headquartered in Bengaluru.

The agent observability space already includes tools such as LangSmith (from LangChain), Arize AI, and Weights & Biases, among others. Traccia's differentiation, as framed by its own HN post, is the addition of runtime governance — the ability to intervene in or constrain agent behavior at execution time rather than analyzing traces after the fact. The company says its documentation is extensive and the platform is designed to be easy to use, though no third-party assessments are available at this stage.

The three-month free trial is the primary near-term offer, and the team indicated openness to feedback from early users. No roadmap items, pricing tiers, or expansion plans were mentioned in the available sources beyond the trial period.