New — Introducing Yaju Software Factory — Try the tool for free
Explore nowPlatform
Functionalities
Use Cases
Deployment
By Industry
Explorations
Initiatives
MEET ORAN
Oran is the third generation of the Yaju Agent Orchestration System, and the one organisations run today. It is the point where agents stopped being something you build and became something you can operate.
NEW RELEASE
The latest Oran version brings integrations, situation awareness and advanced management to the Agent Orchestration System. Agents connect to the tools your teams already use, respond to what is happening around them, and stay under central control as they scale.
Oran 3.6
The latest version: integrations, situation awareness and advanced management.
Oran 3.1 opened the third generation with the question nobody could answer before: what does each agent actually cost? Spend is attributed per agent and broken down by user and by model, and reported in dollars rather than tokens. At this stage it was reporting only, with no limits attached to it yet.
Oran 3.2 turned that visibility into control. Circuit-breaker budgets set hard limits per workspace and team, warn as the limit approaches, block automatically when it is reached, and forecast the rest of the cycle. An overrun is stopped while it happens instead of being explained afterwards.
Two generations came before Oran. Keter was the first runtime, able to execute a single agent and little else, and grew across eight versions into a catalogue with a named owner per agent and a central credential store. Nahar followed across four more. Every version is listed on the versions page.
Oran began with the question nobody could answer before: what does each agent actually cost? Six versions later it attributes spend per agent, stops an overrun before the invoice, scores every run against your own criteria, lowers cost without losing quality, and lets teams share agents instead of rebuilding them.
versions in the Oran generation
versions since the first runtime
generations
capabilities in Oran 3.6
Under the hood
Cost you can attribute
Spend is attributed per agent and broken down by user and model, reported in dollars rather than tokens. The question of what an agent costs finally has an answer, and it arrives before the invoice does.
Budgets that stop an overrun
Circuit-breaker budgets per workspace and team warn as the limit approaches and block when it is reached, with a forecast for the rest of the cycle. Control happens at the moment of spend, not in the monthly review.
Quality enforced on every run
Evals check every run against behavioural, structural, latency and cost criteria, and halt an agent when one fails. Optimizer then lowers spend measured against those same evals, so cheaper never quietly means worse.
Run agents with cost attribution, circuit-breaker budgets and evals enforced on every run.
