One AI API.
Use our stack or build yours.
Create a stack, choose the models it can use, and call it from any OpenAI-compatible client. Astrolabe picks the lowest-cost model in that stack that is likely to answer well.
Client App
OpenAI-compatible agents, apps, and internal workflows.
Astrolabe Cloud
Standard and custom stacks at the gateway edge.
Complex request
Choose from your stack
Fast path
Lower-cost model when it fits
Simple stacks
Each stack is just a named set of models. Put the models you trust in a stack, then call it as astrolabe/auto, astrolabe/coding, or your own stack name.
95% Cost Reduction
Keep premium models available when they are worth it. Everything else can stay on faster, lower-cost models from the same stack.
Workspace private
Custom stacks are private to your workspace. API keys can call your standard stacks and your workspace stacks, but not another workspace's custom stacks.
Choose the stack per request with the OpenAI-compatible model field.
Drop-in Simplicity
Set the model to an Astrolabe stack and send your request.
from openai import OpenAI
client = OpenAI(
api_key="YOUR_ASTROLABE_CLOUD_KEY",
base_url="https://api.astrolabe.run/v1",
)
response = client.responses.create(
model="astrolabe/auto",
input="Route this through my default Astrolabe stack."
)Four defaults, custom when you need it
Use Astrolabe's standard stacks or create a private stack with your own model list.
astrolabe/auto
Managed default route for general AI API requests.
astrolabe/coding
Optimized for code generation, patching, and agent execution traces.
astrolabe/research
Long-context synthesis, source-heavy research, and deep reasoning.
astrolabe/fast
A standard stack for latency-sensitive requests and simple turns.
Ready to create your first stack?
Create a workspace, mint a key, and call your first stack from any OpenAI-compatible client.