One API chooses the lowest-cost model capable of the job.
Run your first requestUse one OpenAI-compatible endpoint from your app or tool.
Capability and cost are checked before a serving model is selected.
See the decision, enforce limits, and pay only for confirmed usage.
Evaluates capability, policy, and price in real time.
Inspect the route, selected model, tokens, latency, and provider-confirmed cost.
Prepaid balance and hard policies stop rejected work before it reaches a paid model.
Change the base URL and model name in OpenAI-compatible tools and SDKs.
Not exactly. ChatGPT is an app you use directly. Astrolabe is the AI connection for apps and tools that ask for an AI provider or API key.
No. Astrolabe gives you a compatible endpoint, an API key, and copy-ready examples. You can start with astrolabe/auto and leave model selection to the router.
Astrolabe evaluates eligible models for each request and selects the lowest-cost option that meets the task. Easy work can use inexpensive models while difficult work can still reach stronger ones.
Yes. Every request records the models considered, the routing checks, the selected model, token usage, latency, and provider-confirmed cost.
Astrolabe works with tools that support an OpenAI-compatible API or let you enter a custom provider URL and API key.
There is no subscription and no token markup. You prepay credit and pay the selected model's hosted provider rate. A purchase fee applies when credit is added.
Fund a workspace, create a key, and connect with one copy-ready example.
