Gateway routing across 50+ models, enterprise SLAs, and cost controls that keep AI under 0.5% of revenue. Embed via the managed agents API - metered credits for everything else.
Managed agents API
Sessions, SSE streaming, and headless runs - persona, allowlist, and console in one surface.
Image API
Gemini and Nova Canvas behind one endpoint; files land in workspace storage.
Workspace agents
Assist and Plan for your team - grounded in storage, gated by approval.
Credit metering
One cent per credit; Insights itemizes every agent, key, and surface.
Endpoint reference, code samples, and console KV live on /managed-agents - not duplicated here.
Pay-as-you-go credits for most teams - see /pricing. Committed contracts when volume and SLAs matter.
One meter. Every agent, every key.
Wallet burn, token volume, and component costs over time - per workspace, per managed agent, per app_id. The same numbers on the invoice.
See credit ratesThis month
$0
0 credits
Balance
$0
$142/day burn
Daily spend (7d)
Spend by surface
Predictable
Cost you control to the credit.
Every turn is metered against a prepaid wallet. Hard budget caps stop spend at your ceiling, prompt and tool-schema caching skips repeat work, and a task-aware router sends each job to the cheapest capable model.
Prepaid credit wallet
Every turn is metered and debited in real time. A funded wallet gates spend before a single token is billed.
Budget caps & alerts
Monthly policy budgets with alerts at 50/80/100% - and an optional hard stop, even on a funded wallet.
Prompt + result caching
Static prefixes and tool schemas are cached, so multi-step turns don't re-pay for the same context.
Task-aware model router
Cheap models for classification and lookups; frontier models only when the task truly needs them.
Not sure which layer you need?
Most teams prove ROI in the workspace first, embed via managed agents when ready, and commit to platform SLAs when volume is predictable.
Workspace credits on /pricing. API depth on /managed-agents.