Inference that speaks the APIs you already use.
Flynn is an OpenAI- and Anthropic-compatible API for open-weight models. Point the SDK you already use at Flynn, keep your code, and either pin a model or let the Flynn router choose one for each request. You pay per token, from prepaid credit.
import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.flynn.digitalfrontier.so/v1",
api_key=os.environ["FLYNN_API_KEY"],
)
reply = client.chat.completions.create(
model="flynn", # the router picks; or pin a model id
messages=[{"role": "user", "content": "Hello, Flynn"}],
)
print(reply.choices[0].message.content)Flynn features
Two wire formats, one key
POST /v1/chat/completions for OpenAI-shaped clients and POST /v1/messages for Anthropic-shaped ones, both behind the same API key. Existing SDKs work by changing the base URL and the key.
A router, or your pick
Send model "flynn" and the router chooses the model that serves each request, or pin any model id from GET /v1/models. A routed request is billed at the list price of the model that served it.
Per-token, prepaid
Input and output tokens are priced separately, in US dollars per million tokens. The price shown is the price billed, and failed requests are not billed. Top up prepaid credit from the console.
Keys you manage
Sign in to the Flynn console and a personal account is provisioned. Mint, rotate and revoke your own API keys; there is no cap on how many you keep active.
Limits you can see
Each account has requests-per-minute and monthly caps. Your values are returned in `limits` on GET /v1/usage, and a refusal is a 429 that names its `reason`.
Changes announced in advance
The stable surface gets 90 days' notice before a breaking change, announced with its date in the public Flynn API changelog.
Build on Flynn.
Sign in, mint a key and send your first request with the SDK you already use.