QUANTA
by Ryshe
Get started
Cut token usage in one line.
Quanta is an OpenAI-compatible gateway. Point your base URL at it, keep using your own provider key, and every request is compressed on the way to the model. No SDK changes, no rewrites.
1. Change your base URL
Use your existing OpenAI client. Set the base URL to the gateway and keep your own key.
from openai import OpenAI
client = OpenAI(
base_url="https://quanta.ryshe.com/api/v1",
api_key="YOUR_PROVIDER_KEY", # your own key, never stored
)
resp = client.chat.completions.create(
model="gpt-5.5",
messages=[...],
)2. Read the savings
Every response carries the savings on that call in its headers.
x-quanta-tokens-before: 8410 x-quanta-tokens-after: 3120 x-quanta-saved-pct: 62.9 x-quanta-mode: inline
Try it now, no key required
This dry-run call returns the compression report without forwarding to a model, so you can see it work right away. Paste it into a terminal.
curl -s https://quanta.ryshe.com/api/v1/chat/completions \
-H "content-type: application/json" \
-d '{"model":"gpt-5.5","messages":[
{"role":"system","content":"You are helpful.\n\n\nYou are helpful.\nBe concise."},
{"role":"user","content":"Summarize this."}
]}'Three ways to adopt
Live
Drop-in base URL
Change one line in any OpenAI-compatible SDK. Nothing else changes.
Self-host
Local proxy
Run the gateway next to your app and route traffic through it. Engine layer is open-standards.
Coming soon
MCP server
Compress tool outputs and context for agent frameworks over MCP.
Accounts, usage metering, and a savings dashboard are in private beta. For now the gateway works with your own provider key and reports savings in headers. The compression engine is built to wrap open-standards and open-source components, which are credited in our notices.