AnonRouter is live

AnonRouter is live today. It is one OpenAI-compatible endpoint in front of more than 200 models, and it keeps no record of what passes through it. You can try it right now without telling us your email address.

Why build another router

A router is a convenience play. You get one API key, one bill, and every model on the market, instead of six provider accounts that each want a credit card. We like that trade as much as anyone. What we did not like is the seat it puts the router in: every prompt from every app you build flows through one company, and the usual practice is to keep some record of it, for debugging or, under some terms of service, for training.

That archive is the part we removed. AnonRouter holds your prompt in memory while the model serves it and lets it go when the response finishes. Billing settles from a single row of metadata: the model, token counts, status, latency, cost, and a timestamp against an opaque account id. If someone demanded our request logs tomorrow, the honest answer is that there is nothing to hand over. The full request path is written up in the privacy model docs.

What shipped today

The core is the API. It speaks the OpenAI Chat Completions surface, so most apps switch by changing the base URL and the key:

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.anonrouter.ai/v1",
  apiKey: process.env.ANONROUTER_API_KEY,
});

const res = await client.chat.completions.create({
  model: "anthropic/claude-opus-4.8",
  messages: [{ role: "user", content: "..." }],
});

Behind it is a catalog of 200+ models across providers. Every route carries a privacy label naming the strongest guarantee it provides: Anonymous, Private, TEE, or E2EE. Routing enforces the labels rather than just displaying them, and you can filter the catalog by the level you need. Moderation stays a model property, so models published uncensored run uncensored.

If you would rather not pick a model at all, send /auto. The router classifies the task in memory, then picks the cheapest capable model inside your privacy boundary, and the classification labels are never stored. There is also a Chat workspace for text models and a Studio for images and speech. Both keep history in your browser rather than on our servers, and Ghost Mode turns even that off.

Accounts that know nothing about you

An account does not require an email address. A private account is a server-assigned username, hashed credentials, and a recovery phrase you write down. We cannot reset it for you, because there is no inbox to send a reset to; that is the cost of an account that nothing identifies. Registered accounts with email exist too, if you prefer having the reset button.

You can pay by card, or with Bitcoin and Lightning if you would rather keep card identity out of it. One honest caveat: the Bitcoin blockchain is public, and payment processors see what processors see. We collect as little as a working service allows, which is not the same as making the rest of the internet forget you. The Privacy Policy spells out the short list of what we do hold.

What it costs

Model usage is charged at the displayed provider rate. The service fee appears separately at credit checkout: currently a promotional 5% with a $0.75 minimum, down from the regular 8%. There is no subscription. You buy credits when you want them, starting at $1, and they sit there until you spend them.

Where to start

Create a key on the API keys page, point any OpenAI SDK at the endpoint above, and pick a model from the catalog. Coming from OpenRouter? The switch is a base URL and a key, and AnonRouter vs OpenRouter covers what else changes. The docs live at docs.anonrouter.ai.