Token Optimisation

Send less avoidable context to your AI provider.

Varion is an OpenAI-compatible optimisation gateway that can compact eligible history, prune irrelevant tool schemas, use exact caching and fall back to passthrough when optimisation is not appropriate.

100,000 free processed input tokens + 50 local test runs after email verification · Provider charges remain separate
One endpoint change

Keep the app. Change the path.

Compatible applications can point their OpenAI client at Varion and use a Varion key. Provider credentials remain under customer control.

OpenAI-compatible chat and Responses endpointsHistory reduction and tool pruning where safeExact request cachingSafe passthrough for unverified workflowsEstimated and paired-verified reporting
from openai import OpenAI

client = OpenAI(
  api_key="YOUR_VARION_KEY",
  base_url="https://api.varion.tech/v1",
)

response = client.chat.completions.create(...)
How billing works

Pay for Varion processing, not a claimed percentage.

One Varion token covers one successful customer input token processed. Failed provider requests deduct nothing. Savings estimates are reporting metrics and do not determine Varion fees.