Latten

How is Latten different from Helicone?

Helicone is built around LLM gateway observability, logging, caching, rate limits, and cost analytics for model calls. Latten is built around company AI governance: who acted, what the AI reached, what it cost, and whether private data exposure is visible without copying values.

See it on a live graph — no signup →

Gateway versus graph

Gateway tools see model traffic well. Latten records boundary crossings across model calls, data domains, external APIs, and agent handoffs, then renders one authority graph.

Cost with an owner

Latten ties model spend to the actor and data domain behind the call, so finance and platform teams can see spend without an owner.

Observe-only by design

Latten does not need to sit in front of requests to create value. The SDK records beside each call and never changes return values.

How it works

  1. 1. Seed actor identity Use withActor at the request or job entry point.
  2. 2. Track model and data calls Use @latten/sdk around the crossings that matter.
  3. 3. Verify real traffic Confirm crossings arrive in the console.
  4. 4. Use the receipt Share what was reached, what it cost, and that PII values did not leave.

Common questions

Is Latten an LLM proxy?

No. Latten is observe-only at launch and sits out of the request path.

Does Latten track non-LLM calls?

Yes. Latten tracks LLM calls, data touches, external APIs, and agent handoffs as crossings.

When is Helicone better?

Helicone is a better fit when the core job is LLM gateway logging and model-call analytics.

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