Cheap by Default, Controlled From Your Phone
Make capability, cost, and safety tradeoffs explicit in routing policy, then use configured remote approval for sensitive work. Arbiter is a local-first working alpha in public binary release preparation.
There is a specific way AI-at-scale wastes money, and it is so common that people stop seeing it. A frontier model, priced for reasoning about hard problems, gets handed a triviality: classify a string, sort a list, answer a yes-or-no. It succeeds, of course. It also costs many times what the work was worth, and nothing in the stack notices. That is undifferentiated compute, the same expensive horsepower spent on every task regardless of how hard the task actually is. It is one of the real bottlenecks in running agents at scale, and it is the one Arbiter exists to remove.
Cost-aware routing under hardening
Arbiter gives operators a place to configure capability, cost, and safety tradeoffs for dispatched work. The routing layer records its decisions, while learned cost optimization remains under hardening for public release.
The coordination itself is the part most orchestrators forget to count. Triage, classification, and chat are real inference. Arbiter can run local orchestration on hardware you own, keeping the decision path local while hardware, power, and connected-provider costs remain yours to choose.
Safety policy belongs beside cost policy
Cost optimization is useful only when it is paired with safety policy. Arbiter supports configured route constraints and approvals for sensitive work, so price need not be the only decision factor.
The whole fleet, without a collision
Arbiter can supervise configured parallel work and present its recorded state in one place. Local IPC is the default control surface; an authenticated HTTP control plane is optional.
Remote approval is configurable
Reading and acting have different risk profiles. When remote approval and TOTP are configured, sensitive actions can require an explicit operator decision before they proceed. Choose the policy that matches your environment and keep the control path on infrastructure you operate.
The conductor of the sovereign stack
Arbiter is the coordination layer of a stack built to keep your compute, your keys, and your decisions on your side of the line. It shares a cost-aware routing vocabulary with Warden, which executes the work across providers, and it shares the local-brain principle with Curator, which keeps memory on hardware you own. The whole design is laid out in The Sovereign AI Engineering Stack.
Arbiter is a proprietary working alpha in public binary release preparation. Follow the public release channel for verified availability updates.