Aurora 3 — pipelines, evals, and routing in one place

Build AI features
without the plumbing

Aurora turns a prompt, a data source, and a test set into a versioned endpoint your team can ship on a Tuesday afternoon.

aurora run support-triage
retrieve14 passages from handbook
routefast model — intent is simple
draftreply generated, tone: calm
evaluatepasses your tone + accuracy checks
done — published to /v1/triage

Teams building on Aurora

NorthwindKelpieVerdantOctavePallasNimbusTesseraHarbour
NorthwindKelpieVerdantOctavePallasNimbusTesseraHarbour

Everything between the prompt and production

Six pieces you would otherwise build yourself, already wired together.

Visual pipelines

Drag steps together and Aurora wires the retries, queues, and fan-out for you.

Model routing

Send each step to the model that suits it.

Versioned prompts

Every change is a diff you can roll back.

Live evaluations

Score outputs against your own test set on every deploy, then compare runs side by side.

Private by default

Your data stays in your own region.

One SDK

Same calls from Node, Python, or the edge.

Three moves, one endpoint

01

Connect

Point Aurora at your docs, database, or warehouse.

02

Compose

Build the flow once in the canvas and name its inputs.

03

Ship

Publish as an endpoint, a job, or a chat surface.

Simple plans

Cancel anytime. Prices shown per seat.

Hobby

For the first prototype.

$0/ month
  • One workspace
  • Community support
  • Local runs
Choose Hobby
Most picked

Studio

For small teams shipping weekly.

$24/ month
  • Unlimited pipelines
  • Shared prompt library
  • Evaluation runs
  • Email support
Choose Studio

Scale

For production workloads.

$74/ month
  • Region pinning
  • Audit log
  • Custom model endpoints
  • Priority support
Choose Scale