chatting — cyrus.is

cyrus.is/chatting

cyrus.is / chatting

ready.

Open settings, add an endpoint, pick a model.

System

enter to send · shift+enter for newline

Archive

Settings

Endpoints

Add endpoint — name

URL

OpenAI-compatible base URL — no trailing slash

API key (optional)

Advanced

Mode

Also switchable from the top bar on desktop

Temperature

Max Tokens

Models — pull / delete

Pull a model (Ollama)

starting…

Installed

Start new session

Save the current conversation to the archive before starting fresh?

Session name

How this works

This is a static HTML file — no server, no database, no proxy on cyrus.is. When you send a message, your browser talks directly to whatever endpoint you configure. Your prompts never touch cyrus.is infrastructure.

Your endpoints, API keys, settings, and archived chats are saved in localStorage on your own device. Nothing is transmitted to cyrus.is — only to the inference server you point it at.

Because this page is HTTPS, the endpoint must also be HTTPS (browsers block HTTPS→HTTP). Use a Tailscale *.ts.net name or a loopback http://localhost tunnel.

If you're running Ollama, set OLLAMA_ORIGINS=https://cyrus.is before starting it, or your browser will block the request with a CORS error. Pull/delete model management appears automatically when an Ollama endpoint is detected.

If someone else visits this page, they configure their own endpoint. Yours is stored only in your browser and is never shared.