Connect your own AI server.

Before you begin

• Start your own AI server and load at least one chat model.
• Confirm it supports GET /v1/models, POST /v1/chat/completions, and streamed chat responses.
• Keep the server private. Writers Studio accepts plain HTTP only for localhost, .local, and private-network addresses; a public endpoint must use HTTPS.

Connect on a Mac

1. Open Writers Studio, then open Settings and select the AI settings.
2. Choose OpenAI-Compatible Local Server.
3. Enter the server's base URL, such as http://localhost:1234 or http://192.168.1.20:8080. You do not need to add /v1.
4. Enter an optional server API key if your server requires one, then choose Apply & Test Connection.
5. Select one of the discovered models and return to the AI panel.

Connect from iPhone or iPad

1. Make the server reachable from the device on your private network. A localhost address on another computer will not work on iPhone or iPad.
2. In AI settings, enable the Advanced Self-hosted option, then choose OpenAI-Compatible Local Server.
3. Enter the server's private-network address and optional server key. Allow local-network access when iOS or iPadOS asks.
4. Test the connection, choose a discovered model, and keep the server protected from untrusted networks.

Troubleshooting

No models appear

Load a chat model on the server and confirm its model-list endpoint is available. Writers Studio does not use a manual model fallback.

The connection test fails

Check the base URL, server port, firewall, and optional key. For iPhone and iPad, confirm both devices can reach the server on the same trusted network.

The URL is rejected

Use HTTPS for a public endpoint. Plain HTTP is reserved for localhost and private-network addresses to avoid sending writing data over an insecure public connection.

A response does not stream

The server must support standard streamed Chat Completions responses. Check its compatibility documentation and server logs.