Latten

How is Latten different from LangSmith?

LangSmith is strongest for developing, tracing, testing, and evaluating LLM applications, especially in the LangChain ecosystem. Latten is for operating company AI safely: actor-level cost, data reach, PII exposure, attribution, and blast-radius visibility across the agents and apps already running.

See it on a live graph — no signup →

Build-time versus operating-time questions

LangSmith helps teams debug and improve LLM behavior. Latten helps teams answer what the deployed AI can reach and what it costs.

Authority matters

For governance, the key object is not only the run. It is the authority chain: who asked, which agent acted, and which service account or data domain was touched.

Receipts for security and finance

Latten produces a receipt that a human can read: reached domains, models, cost, attribution quality, fixes available, and zero-PII-values posture.

How it works

  1. 1. Install @latten/sdk Add observe-only instrumentation to the TypeScript app.
  2. 2. Wrap crossings Track model calls, data access, external APIs, and handoffs.
  3. 3. Review the graph Find broad reach, unattributed spend, and PII exposure.
  4. 4. Keep or fix deliberately Use the graph before changing boundaries.

Common questions

Does Latten run evals?

No. Latten focuses on cost, reach, PII exposure, and governance receipts rather than eval workflows.

Can LangSmith and Latten run together?

Yes. Use LangSmith for trace and eval workflows, and Latten for actor-level cost and reach visibility.

Does Latten need LangChain?

No. @latten/sdk works in TypeScript apps directly and can instrument provider SDK calls.

See a live graph — no signup Instrument your app → All guides