Arrivals vary
One route may repeat in short bursts while another repeats slowly over an hour. The right cache TTL depends on when requests arrive.
qache
The gateway forproduction AI traffic.
Cut spend and hold quality on every LLM request. Start with automatic caching in minutes; qache learns your traffic and pays for repeated work just once. When you're ready, add a custom router that sends each request to the right model for your quality bar. Prompts are processed in memory with zero data retention by default.
Drop-in integration · Works with your existing keys · Zero data retention
Building an agent? Get a gateway key with no sign-up.
Caching today · Routing pilots · Self-hosted available
Quickstart
Send your first request.
OpenAI Chat Completions
/chat/completionsimport OpenAI from "openai";// your provider's base URL, without the https:// schemeconst providerHost = "api.openai.com/v1";const client = new OpenAI({ baseURL: `https://gateway.qache.cc/${providerHost}`, apiKey: process.env.OPENAI_API_KEY!, defaultHeaders: { "X-Qache-API-Key": process.env.QACHE_KEY! },});const completion = await client.chat.completions.create({ model: "gpt-5.4", messages: [ { role: "system", content: "Stable product instructions" }, { role: "user", content: "The current user request" }, ],});
The highlighted lines are the whole change; they send your traffic through qache and turn on automatic caching. Model routing is a separate opt-in; see the docs.
Why cache strategy matters
The best cache strategy depends on both shape and timing.
Shapes differ
Stable instructions mix with changing user input, retrieved snippets, identifiers, and runtime values in different places.
qache adapts
qache learns your traffic patterns and automatically chooses the cache strategy that delivers the best savings and speed for each workload.
Describe your workload (or pick one) and read the cache hit rate.
Start from a workload:
Cache hit rate = how much of each prompt is served from cache. An illustrative model: it shows the shape of the gap, not a guarantee for your traffic.
What qache does
A cache decision layer for your AI traffic.
Observe shape
qache watches request patterns without asking your app to change how prompts are built.
Decide value
The gateway decides the best cache strategy for that workload.
Report savings
Each request can be tied back to cache activity and estimated savings in the dashboard.
Routing for production traffic
Send each request to the right model, not just the biggest one.
Always reaching for a frontier model is the other half of your bill. qache routes each request to the model that clears your quality bar, so you cut spend on the easy ones and keep full strength where it counts.
Tuned to your evals
A router is built around your own evaluation set: what 'good enough' means for your workload, not a generic benchmark. Bring your evals, or we'll help you build them.
Cheaper where it holds, stronger where it counts
Easy requests go to a faster, cheaper model; the ones that need a frontier model still get one. You move cost down without giving up quality.
Caching still applies
Routing runs on top of caching, not instead of it. A routed request still gets automatic cache placement, and you watch realized cost and quality in the dashboard.
Already integrated? Try a platform router (beta) instantly with one header on an aggregator route. See the routing docs. For a router tuned to your own evals, start a pilot with us.
Self-hosted
Run qache in your own infrastructure.
For teams that need prompts and traffic to stay in-house, qache can run entirely inside your own environment, so your data never leaves your perimeter. Available for production deployments on a case-by-case basis.
What stays yours
Keep the pieces developers already rely on.
SDKs
Provider keys
Prompt source
Streaming
Works with the providers you already use
qache is a drop-in replacement for any text-based request to the following providers:
OpenAIAnthropicGoogle GeminiOpenRouterGrok (xAI)Vertex AIAmazon Bedrock
Start in minutes with a free Gateway API Key