AI on Your Own Computer

Filamental AI can run on a model on your own computer, through Ollama or LM Studio, instead of a hosted provider. There's no key, no account and no bill, and nothing you write leaves your machine.

Why you'd use it

Privacy, mostly. If your project holds things you'd rather not send anywhere, a local model keeps all of it on your computer. It also costs nothing per question once the model's downloaded, and it works without an internet connection.

The trade is capability. Local models are smaller than the hosted ones, so they're slower and less dependable on the bigger jobs, such as building a structure from scratch.

What you need

Ollama or LM Studio, installed and running, with at least one model downloaded. Both are free, made by other people, and their own sites cover installing them and choosing a model. We don't supply or host a model ourselves.

Choose a model that can call tools. Tool calling is how the AI reads and edits your project, so a model without it can chat but can't touch anything. Both apps usually say which models support it.

Connecting

1. Make sure Ollama is running, or in LM Studio, start its local server.
2. Click the AI button in Filamental's top bar.
3. Choose Ollama (on this computer) or LM Studio (on this computer) from the provider list.
4. Check the address. Where a key would go, there's a box for the address of the server. We fill in the usual one (http://localhost:11434 for Ollama, http://localhost:1234 for LM Studio), so you'll only need to change it if you've set yours up differently.
5. Click Connect. Filamental lists the models the server has, and you add whichever you like.

From then on it works like any other connection, and you can switch between it and your other models from the dropdown at the top of the chat.

If it won't connect

• "Couldn't reach… Is it running?" The server isn't running, or it's on a different address. Start it, or check the address in the app's own settings.
• "…is running but has no models loaded." The server's up but has nothing to talk to, so load or download a model first.
• The model replies, then fails when it tries to change something. It probably can't call tools, and the error is in its own words. Try one that can.

Exploring documents

Explore a Folder or Document with AI works on a local model too. Documents are read in smaller pieces so they fit, which makes it slower, and a model that can't call tools may fail part by part, with its own message in the report.

What it doesn't do

• It isn't as capable as the big hosted models. Expect to guide it more and check its work.
• We don't choose the model for you. What runs is entirely between you and Ollama or LM Studio.
• It isn't the same as connecting LM Studio over MCP, which is LM Studio using Filamental as a tool from its own window. That's covered in AI Connections & Prompting.