Instrument the boundaries where the app calls models, reads data, and hands work between agents. In the Teeras demo, a customer-discovery sweep can route those crossings through Latten so the run shows what it cost, which model calls happened, and whether private data exposure stayed values-free.
See it on a live graph — no signup →Teeras runs an agentic customer-discovery workflow: read a product, search live communities, judge opportunities, draft replies, and check AEO visibility.
Latten turns that agent run into a receipt: model usage, cost, actor attribution, reached services, and PII types and counts.
One small SDK or installer change, one Teeras sweep, then the Latten graph showing the cost and crossings behind the run.
No. Use sandbox data and demo accounts so no personal information or secrets appear.
No. Latten observes the crossings and returns the wrapped call result unchanged.
It is concrete: an AI app, a real agent run, an instrumentation path, and a cost receipt answer the questions buyers ask.