01
Create a project
Choose an environment and copy the project key from the Refario app.
DocumentationRefario / 2026
From first event to complete trace.
Send the context your application already knows. Refario connects it into a readable run with model, tool, cost, and policy detail.
TypeScriptingest.ts
import { Refario } from "@refario/sdk"
const refario = new Refario({
project: "support-ai",
environment: "production"
})
await refario.run("resolve-order", {
customerId, release, model
})01 / Quickstart
Three steps to the first useful answer.
Start with one real workflow. Validate the path, then add the operational context your team needs.
02
Send a run
Attach workflow, model, tool, customer, and release context to the events you emit.
03
Read the record
Open the trace, confirm cost and latency, then configure alerts and guardrails.
02 / Event model
Only the context that makes telemetry useful.
Refario correlates a small set of clear primitives instead of forcing your AI system into a proprietary framework.
01Run
02Span
03Model call
04Tool call
05Guardrail
06Output
Start with one project
Know what your AI system is doing.
No credit card. Connect real traffic and see the first operational record in minutes.