Multi-provider LLM gateway with built-in usage metering. Track costs per customer, enforce quotas, route requests—one SDK.
We'll reach out to when your spot opens up.
import { ModelRelay, parseSecretKey } from "@modelrelay/sdk";
const mr = new ModelRelay({ key: parseSecretKey("mr_sk_...") });
const text = await mr.responses.textForCustomer(
"cust_abc123",
"You are a helpful assistant.",
"Summarize our Q4 results",
);
// Usage billed to cust_abc123use modelrelay::{ApiKey, Client};
let client = Client::with_key(ApiKey::parse("mr_sk_...")?)
.build()?;
let text = client.responses()
.text_for_customer(
"cust_abc123",
"You are a helpful assistant.",
"Summarize our Q4 results",
)
.await?;
// Usage billed to cust_abc123key, _ := sdk.ParseAPIKeyAuth("mr_sk_...")
client, _ := sdk.NewClientWithKey(key)
text, _ := client.Responses.TextForCustomer(ctx,
"cust_abc123",
"You are a helpful assistant.",
"Summarize our Q4 results",
)
// Usage billed to cust_abc123Your users make AI requests. ModelRelay tracks costs per customer. You keep the margin.
Every request is attributed to a customer. Export usage for invoicing or sync to Stripe.
Define free/pro/enterprise tiers with different models and rate limits per tier.
Route to Anthropic, OpenAI, or xAI. Automatic failover when a provider is down.
Set spend limits per customer. Requests fail before they cost you money.
Every response includes token counts and cost. No waiting for monthly invoices.
Define parallel pipelines with workflow.v0. Fan-out, join, tool calling.