# How do I run my bots on a local model?

Source: https://help.locul.ai/articles/run-locul-with-a-local-model

In **Settings** > **AI model**, choose **Your own server**. Locul looks for a model server on this computer and your network, and nothing you send leaves your network.

## Before you start

You need a model server. Locul finds Ollama, LM Studio, llama.cpp, vLLM and Jan. If you have none, install Ollama or LM Studio. A recent Apple Silicon Mac can run small open models. You don't need a special graphics card.

## Use a model on this computer

1. Click the **Settings** gear at the bottom of the left rail.
2. Under **Bots**, click **AI model**.
3. Under **Default for all bots**, choose **Your own server**.
4. Under **Found on this computer and your network**, click **Use** next to your server. Click **Look again** if you just started it.
5. Under **Model**, pick a model.
6. Under **Connection check**, click **Test**. The first test can take a minute while the model loads.

When the test says **Ready**, your bots use this model.

If Ollama is on this computer, Locul may suggest the best model for it. Click **Download** to get it. It runs only on this computer.

## Use your firm's server

1. Under **Add a server by address**, type the address, for example `ai.yourfirm.lan` or `10.0.0.5:11434`.
2. If it needs a key, click **The server needs a key** and paste the **Access key from your IT team**.
3. Click **Connect**, pick a model, then click **Test**.

Plain http addresses work only inside your own network.

## Good to know

- Bots need models that can use tools. Chat-only models are left out of the list. Click **Why are N models missing?** to see which.
- **Memory window** sets how much of a long chat the model keeps in mind. **Auto** is fine for most people.
- Set a **Backup model** to answer when your main model can't, for example when the server is off. The chat tells you when it does.
- If your server is on your network, you can click **Build them on [server name] too** so your memories are built there as well.

## Troubleshooting

**"No model server running on this computer."** Start Ollama or LM Studio, download a model such as Qwen3 8B, then click **Look again**.

**The test says Not ready.** Make sure the model is downloaded and the server is running, then click **Test again**.

**"Your admin has not allowed your own servers."** Your firm controls this option. Ask your admin.

## Related articles

- [Choose the AI that runs your bots](/articles/choose-the-ai-that-runs-your-bots)
- [Cloud and local: where Locul runs](/articles/cloud-and-local-where-locul-runs)
- [Large teams, SSO and on-premise](/articles/large-teams-sso-and-on-premise)