Skip to main content
Providers & Tools

Local models

Chat with models that run on your PC through Ollama or LM Studio, or connect any OpenAI-compatible server.

Written By Daniel Nguyen

Ollama and LM Studio run AI models on your PC, with no API key. BoltAI for Windows connects to them while they're open. Download models in Ollama or LM Studio.

Use Ollama

  1. Install Ollama from ollama.com and open it.
  2. Download a model. For example, in Terminal, run ollama pull llama3.2.
  3. In BoltAI, open Settings › AI Providers and click Add provider.
  4. Click Ollama. Keep Base URL as http://localhost:11434, and click Continue.
  5. In any chat, open the model picker and pick an Ollama model.

Use LM Studio

  1. In LM Studio, download a model and start its local server.
  2. In BoltAI, open Settings › AI Providers and click Add provider.
  3. Click LM Studio. Keep Base URL as http://localhost:1234, and click Continue.
  4. In any chat, open the model picker and pick an LM Studio model.

See models you downloaded later

Open Settings › AI Providers, click Ollama or LM Studio, and click Refresh Model List.

Use a server on another computer

If Ollama or LM Studio runs on another computer on your network, set Base URL to that computer's address, for example http://192.168.1.20:11434. The server must be set up to accept connections from other computers. See its own documentation.

Connect a custom server

Use Custom Server for any server that works like the OpenAI API, such as a company gateway or a self-hosted model.

  1. Open Settings › AI Providers and click Add provider.
  2. Under Other providers, click Custom Server.
  3. Enter the server's Base URL, for example https://api.example.com/v1.
  4. If the server needs a key, paste it into API key.
  5. Click Continue.

BoltAI shows the models the server lists.

BoltAI can't reach Ollama or LM Studio

If a reply fails with Couldn't reach Ollama at http://localhost:11434, or the same for LM Studio:

  1. Check that Ollama or LM Studio is open, and that LM Studio's local server is started.
  2. Open the provider in Settings › AI Providers and check Base URL.
  3. Click Retry on the failed reply.

If BoltAI says Ollama has no models yet, download one in Ollama first, for example with ollama pull llama3.2.

Was this helpful?

Still need help? Ask the team