Local AI models
Route inference directly into their Amazon Bedrock
Or run open-weight models on your customers' GPUs.
POST ai.alien.dev/customer-1/v1/chat/completions{ "model": "byo/claude-opus-5", "messages": [ … ]}- Customer 11,285 requests todayAmazon Bedrockus-east-1 · billed to them
- Customer 2913 requests todayVertex AIeurope-west1 · billed to them
- Customer 3406 requests todayAnthropicbilled to them
Prompts stay under their security terms. Usage counts toward the committed spend they already have. Your product stops paying for inference.
Enterprises want sensitive prompts on their own AI provider.
Their zero data retention becomes yours.
Every term they negotiated with their AI provider covers your requests: zero data retention, no training, their region.
- Zero data retentionNothing stored after the request
- Never used for trainingTheir enterprise agreement
- Stays in us-east-1Their approved region
Cloud identity instead of API keys.
On AWS, Google Cloud, and Azure the customer grants a role, not a key. Nothing to paste, store, or leak.
Forward the inference bill.
SaaS margins, even with AI in the product.
When each customer’s AI provider bills them directly, you no longer pay for inference at all.
Inference moved to each customer’s AI provider.
Let them pay with credits they already committed.
They would rather your product draw down the millions they committed to their AI provider than pay you separately for the same inference.
Keep your OpenAI or Anthropic client.
Add the customer ID to the requests you already send.
Customers who connected an AI provider get their requests run there. Everyone else keeps going through your current gateway, like OpenRouter.
Add local AI models to your product import OpenAI from "openai"
const ai = new OpenAI({
- apiKey: process.env.OPENAI_API_KEY,
+ apiKey: process.env.ALIEN_AI_KEY,
+ defaultHeaders: { "X-Alien-External-ID": customer.id },
})
const response = await ai.chat.completions.create({
- model: "gpt-5.6-sol",
+ model: "byo/gpt-5.6-sol",
messages: [{ role: "user", content: "Summarize this document" }],
})
Everything around the request.
White-labeled onboarding.
A guided setup flow in your branding, with the exact steps for each provider.
See supported providersConnect your AI provider
Choose the AI provider that will serve model requests for your organization.
- You grant
- An IAM role in your AWS account
- Set up with
- CloudFormation
- Required model
claude-opus-5
Your credential stays with Alien. Your product never receives it.
A heartbeat on every customer.
Alien checks every connection continuously, so you know before the customer does.
What model status meansLogs for every request, across every customer.
Status, customer, model, provider, latency, tokens. Never the prompt.
Diagnose requestsOne line per request. Prompt and response bodies are never recorded.
Cost by customer.
What each customer’s usage costs on their own AI provider.
Inspect usageYour bill for the same period: $0.
Run it in your cloud.
Install it yourself and no request passes through Alien.
Connect your first customer’s AI provider.
Enable one model, create a setup link, and send a request through a provider you control. Then offer it to the customer who asked.