Outborn Agent Platform · Manual
Outborn AI runtime
Use Outborn AI as the managed AI runtime module inside Outborn Agent Platform, alongside provider subscriptions, APIs, local models, routing, fallbacks, and usage controls.
Applied guidance
Best practices for outborn ai runtime
Recommended
- Choose a runtime source according to ownership, available models, and cost boundary.
- Check failure and fallback behavior before relying on multiple sources.
Avoid
- Do not combine Hosted AI credits, provider subscription allowance, and provider API billing into one quota.
How it fits together
01
AI runtime overview
OutbornAI can route work through distinct runtime sources. The selected source determines where model access comes from and which allowance or billing boundary applies.
02
Hosted AI
Uses Outborn-hosted model access and Outborn plan or credit allowance.
03
Provider Subscription
Uses a connected user/provider account subscription allowance when that provider and authentication method are supported. It is not BYOK and it is not Provider API.
04
Provider API
Uses provider API credentials. Usage and billing follow the provider's API terms and account.
05
Model selection
Select models from the runtime sources available to the organization or agent. Availability can differ by provider, account, region, policy, and plan.
06
Multiple providers
An agent may have more than one runtime source. Configure the sources in the order you want them considered.
07
Routing priority
Priority expresses the preferred order of eligible runtime sources. It does not merge the allowance or billing of those sources.
08
Automatic routing
Automatic routing selects among eligible configured sources according to runtime policy and availability rather than forcing one provider for every request.
09
Fallback behavior
Fallback can move to another eligible configured source when a preferred source cannot serve the request. Authentication and policy boundaries still apply independently to each source.
10
Reasoning and effort
Where supported, reasoning or effort controls change how a model spends inference effort. Available values depend on the selected provider and model.
11
Usage and limits
Always read usage in the context of its source: Outborn allowance, Provider Subscription allowance, or Provider API billing.
Related documentation