Invite-only access · OpenAI-compatible

One API for leading AI models — with lower costs on eligible models.

Use one OpenAI-compatible API to access the models available to your account, switch without rebuilding your integration, and track every request, token, and charge in one place.

One API key OpenAI-compatible Leading model families Prepaid balance Request-level cost visibility

How a request flows — product illustration

Your request Your app calls one OpenAI-compatible endpoint.
Astrodyne API One key, one base URL. The request is authenticated and checked against your balance.
Eligible models Astrodyne confirms the model you named is available to your account.
Selected model The model you asked for runs the request.
Result The response comes back in the OpenAI-compatible shape, with a request ID.
Usage and cost The request, its tokens and its charge appear in your account.

You choose the model. Astrodyne does not substitute a different one.

How it works

Three steps, then you are building.

01

Connect once

Point your existing OpenAI-compatible client at the Astrodyne base URL and use an Astrodyne API key. Where compatibility is verified, that is the whole change — the OpenAI SDK works as-is.

02

Choose a model

Name the model you want in the request. Your account sees exactly the models it can call, in the console and at GET /v1/models.

03

Track usage and cost

Every request returns an ID you can look up. Tokens, status and the charge for that request appear in your account, against a prepaid balance you top up in advance.

Why Astrodyne

One integration instead of one per vendor.

One integration

Write against a single OpenAI-compatible surface instead of maintaining a client, an auth scheme and a billing relationship per vendor.

Multiple model families

Reach leading families through the same endpoint, and change which model you call without rebuilding the integration around it.

Lower cost on eligible models

Eligible models are priced below the model vendor's published direct API rate. Eligibility is per model, and the comparison is shown per model — never as a single site-wide number.

Simpler billing

One prepaid balance in US dollars. You are charged for usage that has settled, at the price shown for that model.

Visibility into requests

Look up any request by ID and see its status, tokens, latency and charge. No estimating from a monthly total.

Fewer vendor accounts

One account, one key, one balance — instead of separate signups, keys, invoices and spend limits to reconcile by hand.

Useful for AI agents, apps, automations, prototypes and production workflows.

Cost

Lower prices on eligible models.

Eligibility is decided per model, not across the catalog. Where a model is eligible, we show the dollar figures side by side so you can check the arithmetic yourself.

Pricing comparison unavailable

We only publish a price comparison when there is an approved public price for the model and a current published vendor price to compare it against. That is not the case right now, so we are not showing one.

How we compare prices · See the model catalog

For agent builders

Agents make a lot of calls.

A multi-step agent turns one user action into dozens of model calls. That makes per-request visibility and a hard spending limit far more useful than a monthly invoice.

Coding agents

Plan, edit and test loops that call a model many times per task. Inspect any single step by request ID when a run goes wrong.

Research agents

Long chains of retrieval and summarisation, where the cost is in the volume of calls rather than any one of them.

Customer-support agents

Classify, retrieve and draft. Keep the model choice separate from the integration so you can change it later.

Content workflows

Batch generation and rewriting, with the token counts and charge for each run recorded against your balance.

Internal automations

Back-office jobs that run unattended. A prepaid balance is a spending limit that cannot be exceeded by a runaway loop.

Multi-step agent systems

Several agents, several models, one integration and one balance — instead of a vendor account per component.

Read more about Astrodyne for agents →

Integration

Change the base URL and key. Keep your code.

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.astrodyne.ai/v1",
    api_key=os.environ["ASTRODYNE_API_KEY"],
)

resp = client.chat.completions.create(
    model="YOUR_MODEL_ID",
    messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)
Set ASTRODYNE_API_KEY in your environment. Base URL https://api.astrodyne.ai/v1.

YOUR_MODEL_ID is a placeholder — replace it with a model ID from your console. No model is currently published for public examples. See the compatibility guide for supported fields and known differences.

FAQ

Common questions.

What is Astrodyne?

One OpenAI-compatible API for leading AI models, with API keys, a Playground, a model catalog scoped to your account, a prepaid balance, and request-level usage and cost visibility.

Who is it for?

Developers, small teams, agencies and people building AI agents who want one model integration and one bill instead of one per vendor.

Is the API OpenAI-compatible?

Astrodyne implements an OpenAI-compatible Chat Completions API. Point an OpenAI-compatible client at the Astrodyne base URL with your key. The compatibility guide lists supported fields and known differences.

Which models can I use?

The models available to your account, which are listed in your console and at GET /v1/models. Availability is per account, so what you see there is what you can call.

Does Astrodyne pick a model for me?

No. You name the model in your request and Astrodyne runs that model. It does not substitute a different one.

How does billing work?

You add funds to a prepaid balance in US dollars and are charged for usage that has settled, at the price shown for that model. Each request's charge is visible individually.

How do the lower prices work?

Eligible models are priced below the model vendor's published direct API rate. Eligibility is per model, and we publish a comparison only when there is an approved public price and a current published vendor price to compare against.

Can I test in the browser?

Yes. The Playground sends real requests through the same API path your code uses, using a key you supply that is held in memory only.

Do you say which vendors run the models?

No. Customer-facing surfaces present a neutral model catalog and sanitized request data.

Is streaming supported?

Yes, over server-sent events, through the same Chat Completions endpoint.

Can I sign up today?

You can create an account and explore the console. Access to send requests is invite-only while we roll out.

Get started

Start building with one API.

Create an account and set up your integration today. Sending requests is invite-only while we roll out, so access is granted in small groups — you can build against the API, generate keys and explore the console while you wait for yours.