Allocate documentation

Build on the Allocate inference platform.

Allocate gives every company its own Private Inference Cloud: a dedicated gateway scope, data plane, weights, and meter. You bring traffic and data; the platform runs inference, agents, and training inside a boundary you control.

Start here

• Make your first inference call in under five minutes.
• Install the agent skill so your coding agent knows the platform.

The platform

• Gateway: one endpoint for every model, metered from the first token.
• Routes: name the intent in your code, pin the model on the platform.
• Models: the full catalog behind one key, frontier and open weight.
• Keys: per-organization credentials, scoped as narrowly as you need.
• Meter: every request accounted for as it happens.
• API reference: the full wire contract, generated from the gateway's spec.

The planes

• Private Inference Cloud: your boundary, your region, your policy.
• Cloudlight: the default low-latency plane.
• Cloudheavy: region-pinned compute and storage for residency.

The learning plane

• Agents: hosted agent harnesses with human-gated write paths.
• RecursiveDB: your experience database, compacting into a model you own.
• Training: from frontier APIs to fine-tuned open weights.