Manifesto
Model choice should be infrastructure, not product strategy.
AI is becoming a market of thousands of models, providers, regions, prices, and policies. Your users should not have to understand that market—and your product should not be trapped inside one corner of it.
Breadth without dependence.
A useful intelligence layer cannot end at six hand-picked models. It should connect directly to every major provider, keep discovering what comes next, and normalise that changing market behind one stable interface. No upstream aggregator deciding what we can offer, how we route, or where our margin comes from.
Evidence beats reputation.
A famous model is not automatically the right model. New models should earn traffic against real tasks, and routing should choose the most cost-effective option proven good enough for the work. When the stakes rise or confidence falls, the quality floor rises with them.
Optimise the conversation, not the isolated call.
A low-priced next request can make the whole thread more expensive if it throws away a warm cache. Augur treats cache affinity, switching penalties, latency, and quality as part of the same economic decision. The unit of optimisation is the outcome, not a headline token price.
Governance belongs in the route.
Provider, region, data handling, capability, latency, and budget rules should be enforced before a request leaves your product. The same layer that selects a model is the right place to make those constraints unavoidable.
Cost truth is a system of record.
Provider-reported usage, cached tokens, the exact price version, latency, policy, and routing reason belong on every receipt. You should be able to explain what ran, why it ran, what it cost, and what would have happened under a different policy.
One plane, many products.
Augur can power a chat, an agent, an internal workflow, or another company’s application. The interface stays stable while providers and models keep changing underneath it. That is how model choice becomes infrastructure.
Why Augur?
To augur something is to predict its outcome from the signs around it. An ancient Roman augur read the flight of birds, the weather, and the lay of the land before a decision was made. The augur did not decide the policy; they interpreted the signs so the right path could be chosen.
Augur does the same for AI work. It reads the request, the evidence, the policy, the cost of the moment, and the state of the conversation, then chooses the route that can deliver the outcome. No default-provider bias. No premium model by habit. Just a decision you can inspect and defend.
Ready to see it?
Open the demo