Msty Studio 3.x · Desktop guide

External Model Servers

Connect models served by Ollama, LM Studio, or another compatible endpoint.

Connect an existing server

  1. Start the model server and load or install the models it should serve.
  2. In Msty, open Settings > Model Providers > Add Provider.
  3. Select the matching provider type, such as Ollama Remote, LM Studio, or an OpenAI-compatible provider.
  4. Enter its endpoint and any required token or API key.
  5. Fetch or select the available models, then save the provider.

The server must remain running and reachable while you use its models. Use the address reachable from the computer running Msty; localhost refers to that computer.

Endpoint details

  • Ollama: enter the service endpoint, commonly http://localhost:11434 for a local installation.
  • LM Studio: use the server's API endpoint, commonly http://localhost:1234/v1. If its API authentication is enabled, enter the API key too.
  • OpenAI-compatible servers: use the base URL expected by that server, often including /v1.

If model discovery is unavailable, add a custom model using the exact model ID recognized by the server.

Connect Msty Nexus

Settings > Model Providers also offers a Msty Nexus connection. Open Nexus, choose Connect in Studio, approve the displayed code in Nexus, and return to Studio. Studio then loads the models available through that connection.

Import Models
Use a local GGUF file or a compatible safetensors model directory.
Model Parameters
Adjust response settings, thinking effort, and reusable parameter presets.