Use the models you already run.

Your Machine connects Noema to an existing local model server. Noema does not install a server or start a model for you.

Connect a server

1. Start the server and make the model available in its own application.
2. In Noema, choose Your Machine and the connection type.
3. Enter the local server address the application provides.
4. Connect, choose an available model and send a short prompt.

Choose Ollama or Local API in the connection controls. Local API supports compatible servers that expose a model list and chat completions.

Server applications with authentication

Some local applications need an API key and browser access to be enabled. In the Local API controls, the optional key is kept only for the session and the selected server address. It is not saved to a Noema cloud account.

If a server works in its own interface but not in Noema, check whether it allows requests from your browser page. Follow that application’s instructions rather than disabling authentication.

Another machine on your network

You can use a model running on your computer from another computer, phone or tablet on the same network. A loopback address such as localhost or 127.0.0.1 always refers to the device running the browser, not your other computer.

1. Allow network access in your model server’s settings. It must listen on the computer’s network interface, not only loopback.
2. Keep both devices on a network that lets them reach each other. Guest networks may isolate devices.
3. Enter the server’s private address and port. For example, http://192.168.1.20:11434 for Ollama, or https://studio.local:8888/v1 for a compatible Local API.
4. Allow the Noema page’s origin in the server’s browser access settings. Keep server authentication enabled if it requires a key.
5. Connect, select a model and test a short prompt.

Ollama already supports a configured network address. Local API controls also accept private IPv4 addresses, IPv6 unique-local addresses and local names ending in .local or .home.arpa.

Requests travel directly from your browser to the selected server, without passing through a Noema cloud proxy. This is processing on your own server, not necessarily on the phone itself. A model that calls a cloud service still uses that service.

Browser connection requirements

An HTTPS Noema page can be blocked from calling a plain HTTP server. Some browsers offer a local-network permission; others require HTTPS on the server with a certificate your device trusts. HTTPS on your local server is the most compatible setup. Do not disable browser security or server authentication to work around a failed connection.

If the connection fails, check the address, port, firewall, browser permission, HTTPS certificate and the server’s allowed origins (CORS). Noema cannot override these settings on another device. Using a server’s own interface successfully does not establish that it permits requests from Noema.

Model support

The connected model determines whether images or other inputs are supported. Start with plain text to check the connection before trying a more complex request.