Aerostack
Aerostack
AI Proxy is coming soon — the features below show our vision for this product.
Coming Soon

APIs cost money. This one earns it.

The AI Gateway
that pays for itself.

Ship an AI API with built-in RAG, content moderation, multi-provider fallback, and per-consumer billing. Your users get an OpenAI-compatible endpoint. You get revenue.

RAG PipelineModerationMulti-LLMToken BillingConsumer KeysBYO-JWT
//Pipeline

Build your pipeline. Toggle each stage.

Every stage is optional except LLM. Enable what you need — skip what you don't.

RAG

Enabled

Retrieve relevant context from your knowledge base and inject it into the prompt.

Upload documents — auto-chunked and embedded
Vector search finds relevant context for each query
Context injected as a system message before the user's question
Configurable similarity threshold and top-k results
//Consumer keys

Your API. Their keys.

Each consumer gets a unique API key. You control access, track usage, and bill per token — all automatic.

consumer-call.sh
// Your consumer calls your API — not OpenAI's
curl -X POST https://gateway.aerostack.dev/my-api/v1/chat/completions \
  -H "Authorization: Bearer ask_live_7f3a9c2e4b1d..." \
  -d '{
    "messages": [{ "role": "user", "content": "Summarize Q4 revenue" }],
    "stream": true
  }'

// OpenAI-compatible response — any SDK works
// Your RAG context, moderation, and billing all happen transparently

One key per consumer

Issue API keys with ask_live_ prefix. SHA-256 hashed — raw key shown once. Revoke or regenerate anytime.

Token wallet billing

Each consumer has a token balance. Every request deducts tokens used. Set hard limits to prevent overspend.

OpenAI-compatible endpoint

Your consumers use the same /v1/chat/completions format they already know. Any OpenAI SDK works out of the box.

BYO-JWT — bring your own auth

Already have an auth system? Validate your own JWTs against your JWKS endpoint. No migration needed.

//Reliability

Never go down. Automatic fallback.

Configure a chain of LLM providers. If your primary fails, the next one picks up — automatically, per status code.

Primary: Claude Sonnet

active
fails →

Fallback 1: GPT-4o

on 429, 503

standby
fails →

Fallback 2: Gemini Flash

on any error

standby

Per-status-code routing

Route 429 (rate limit) to one provider, 503 (outage) to another. Fine-grained control.

Transparent to consumers

Your API key and endpoint stay the same. Consumers never know a fallback happened.

Provider-agnostic format

Your consumers send OpenAI format. You can run Claude, Gemini, or Groq behind it.

//Why Aerostack

Not another API gateway.

Traditional gateways route traffic. This one adds intelligence.

Kong / AWS API GWOpenRouterAerostack
AI-Native Features
Built-in RAG pipeline
Content moderation stage
Pre/post processing hooks
Multi-provider fallback chains
Billing & Access
Per-consumer token wallets
Consumer API key provisioning
BYO-JWT (your own auth)
Usage-based hard limits
Platform
Edge-deployed (300+ locations)
OpenAI-compatible endpoint
Marketplace monetization
Full backend platform included

Launch your AI API.
Start earning per token.

RAG. Moderation. Fallbacks. Billing. All configured — not coded.