GET /v1/models
List the public model line with capabilities, lifecycle status and availability — never reserved or retired models.
Ottili AI Platform
One AI platform behind every Ottili product — and your own apps.
Ottili AI Platform is the shared AI capability layer for Ottili ONE. It gives every product — Ottili AI, Ottili Coder, Ottili LD3 and Ottili Cloud — the same models, credits and controls, and exposes that same capability to developers through an OpenAI-compatible API.
Platform, not a single product
A shared foundation, not another chatbot.
Ottili AI is one product built on this platform — the assistant, missions and approvals you use in the workspace. The platform underneath is the part every product shares: model access, routing, credits, usage controls and reliability. Developers reach the same models through api.ai.ottili.one, so your own apps speak the same OpenAI-compatible contract as Ottili's own products.
Model access & routing
Models, routing and provider abstraction.
A small, curated model line with an adaptive router, served through one provider-abstraction layer. Routing fails closed: an unavailable model is reported, never silently downgraded to a paid or mock provider. Bring-your-own-key is explicit and company-scoped.
Model access
BetaOttili Vale 1.2 (efficient), Ottili Cairn 1.2 (premium) and Ottili Auto (adaptive router) — one stable set of public model ids.
Model routing & provider abstraction
LiveOne routing layer maps public model ids to backing providers. Upstream provider credentials never cross the public boundary.
Adaptive router
BetaOttili Auto selects the efficient model for fast everyday work and the premium model for heavy or agentic work, per request.
Company-scoped BYOK
BetaBring-your-own-key credentials live only in the approved credential store — never returned, logged or shared across tenants.
Reserved flagship
PlannedOttili Spire 1.2 is reserved for a later release. It has no public backing model and is never served until launched.
Public model line: Ottili Vale 1.2, Ottili Cairn 1.2, Ottili Auto. Ottili Spire 1.2 is reserved.
Ottili Vale 1.2
AvailableEfficient tierEfficient tier
The efficient everyday model for chat, drafting and fast coding tasks.
Ottili Cairn 1.2
AvailablePremium price-performance tierPremium price-performance tier
The premium price-performance model for advanced agentic and reasoning-heavy software work.
Ottili Auto
AvailableAdaptive routerAdaptive router
The adaptive router. It selects Ottili Vale for fast everyday work and Ottili Cairn for heavy or agentic tasks, per request.
Ottili Spire 1.2
Coming soonFlagship tierFlagship (reserved)
The reserved Ottili AI flagship. Not yet available — planned for a later release.
OpenAI-compatible API
The same contract your tooling already speaks.
The public surface is OpenAI-compatible and deterministic. One API key, company context resolved before any request, and no private detail in the response.
POST /v1/chat/completions
Chat and streaming completions with the same request shape, finish reasons and normalized error envelope as OpenAI.
POST /v1/responses
A first-class Responses API: provider-independent input, output items and correlated usage — not a thin chat wrapper.
API keys & SDKs
Issue company-scoped keys from the console and call the platform from any OpenAI-compatible SDK.
Company credits & usage controls
Credits, quotas and tenant isolation.
Usage is metered against a company wallet. Every billable request takes a worst-case reservation before work begins, settles on success and releases on abort — so retries and replays never double-charge.
Company credits
A single company wallet backs AI usage across every product and the public API. Spend is metered per request.
Reserve → finalize → release
Idempotency keys and a preflight reservation settle exactly once. Cancellations release the unused reservation.
Rate limits & quotas
Per-company, per-project and per-user limits enforce fair usage and protect the shared platform.
Tenant isolation
Every request resolves to an authenticated key or session plus company context before dispatch. No cross-company data.
Routing & cost controls
How every request is routed, charged and capped.
The same control layer that powers Ottili AI, Coder, Ottili LD3 and Cloud governs how each request is routed, cached, charged and capped. The rules below are product-level: no internal provider names, no credentials, no private Unified API detail (website audit P0.2).
Provider routing
LiveOne routing layer maps public model ids to backing providers by capability and tier. The adaptive router (Ottili Auto) picks the efficient model for fast work and the premium model for heavy or agentic work, per request.
Fail-closed fallbacks
LiveWhen a model or provider is degraded, the request is reported — never silently downgraded to a paid or mock provider. You always know which model served the response.
Circuit breakers
LiveA failing provider or route is isolated behind a circuit breaker so one degraded dependency cannot take down the shared platform. A tripped circuit returns a normalized error instead of hanging.
Cache use
BetaRepeated prefix tokens are cached per company, so steady-context workloads cost less and run faster. Cache is scoped to your company and never shared across tenants.
Credit charging
BetaEvery billable request takes a worst-case reservation before work begins, settles on success and releases on abort. Idempotency keys mean retries and replays never double-charge.
Hard caps
LivePer-company, per-project and per-user hard caps stop spend at a ceiling regardless of retries. Exceeding a cap returns a structured, machine-readable error — no silent overage.
Wallet recharge
BetaCompany wallets are topped up through the active plan or purchased credit packs. Spend is metered per request against the wallet, so usage maps cleanly to what was paid for.
Company wallets
BetaA single company wallet backs AI usage across Ottili AI, Coder, Ottili LD3, Cloud and the public API. Spend is company-scoped and isolated — no cross-company charges.
Caching, reliability & structured output
Built for production workloads.
The platform is designed for predictable, observable inference: prompt caching, structured outputs, tool calling and fail-closed reliability.
Prompt caching
Repeated prefix tokens are cached so steady-context workloads cost less and run faster.
Structured outputs
Request JSON-mode output for deterministic, parseable responses your code can rely on.
Tool calling
Function calling with a validated allow-list — bounded tool count, names and payload sizes.
Fail-closed reliability
Normalized errors, idempotent retries and streaming that settles exactly once on disconnect.
Built into every Ottili product
One platform, shared by the whole ecosystem.
The platform is not a separate surface you adopt on its own. It is the AI layer Ottili products already run on — and the same layer your developers can call directly.
Ottili AI
The assistant, missions and approvals are built on the platform's models, credits and controls.
Ottili Coder
Coder uses the platform for agentic coding, planning and code review across local, cloud and hybrid runs.
Ottili LD3
Ottili LD3 drafts, reviews and publishes content with the platform's models and structured outputs.
Ottili Cloud
Cloud runs platform inference and Coder workloads on one Ottili-owned control plane.
Availability
Available in Early Access.
Ottili AI Platform is available through the Ottili Early Access program. The public OpenAI-compatible API, the model line and company credits are reachable today; capabilities marked beta are in active rollout. No dates are promised here.
Build on the Ottili AI Platform
Issue an API key, read the public reference, and call the same models that power Ottili ONE.