Decisions API alternatives you can call today

OpenAI's Decisions API is in limited preview with no public endpoint. If you need a finite-answer decision call now — classification, routing, agent next action — these are the real options, including ours.

OpenDecisionsAPI /v1/decisions

Live · this site

The endpoint behind our playground: TypeSafe's Jev System One model answering bounded questions. Input-token billing, per-option probabilities, idempotency keys. A drop-in stand-in while you wait for OpenAI access.

Endpoint: POST https://opendecisionsapi.com/v1/decisions (alias /v1/systemone)
Types: choice · noul (yes/no) · score
Model: jev-latest (jev-1.13.0) — TypeSafe, not OpenAI
Docs: /docs · pricing

TypeSafe Jev — System One direct

Live · upstream provider

The upstream our endpoint calls. If you'd rather go straight to the source, TypeSafe publishes its own System One API with the same question types.

Types: noul · choice · score
Pricing: $42 / 1B input tokens ($0.042 / 1M), output free
Docs: docs.typesafe.ai · typesafe.ai

OpenRouter Decisions API (alpha)

Alpha · aggregator

OpenRouter exposes a Decisions-shaped alpha route backed by the Jev model (typesafe/jev-1.13). Useful if you already hold OpenRouter credits. Not an OpenAI product despite the similar name.

Endpoint: POST https://openrouter.ai/api/v1/api/alpha/decisions
Model: typesafe/jev-1.13
Docs: OpenRouter alpha docs

LLM Gateway — /v1/systemone

Live · third-party gateway

A community-built decision API that “returns values, not text”, also implementing the System One shape. Covered in a DEV Community write-up.

Endpoint: POST /v1/systemone
Docs: dev.to write-up

OpenDecision — local, open source

OSS · runs locally

A pip-installable local engine using natural-language inference to pick from your finite answer set. No API key, no per-call cost, no data leaving your box — at the price of running the model yourself.

Install: pip install open-decision-ai
Links: PyPI · project docs

Structured outputs / JSON mode

Works everywhere · the old pattern

Constrain any chat model with a JSON schema and parse the result yourself. Every provider supports it, but you own the prompt engineering, the parsing, the retry loop — and you don't get calibrated per-option probabilities.

Shape: chat completions + response_format JSON schema
Trade-off: zero new dependencies; more glue code, weaker guarantees

Independent developer resource. Not affiliated with or endorsed by OpenAI or TypeSafe. “OpenRouter Decisions” and “OpenDecision” are unrelated to OpenAI despite the naming.