Run More AI fully offline with local models (Ollama & LM Studio)
Point More AI at a model running on your own hardware and nothing ever leaves your machine — no API bill, no internet required. Here is how to set up Ollama or LM Studio in minutes, and what to expect from local models in each mode.
The More AI team · Updated July 16, 2026
Why run a model locally
Every cloud model — even through your own API key — means your prompt travels to someone’s server. A local model removes that last hop: the model runs on your own hardware, More AI talks to it over localhost, and your conversations never touch the network at all. That’s the setup for sensitive documents, air-gapped machines, or simply working on a plane.
The second reason is cost: a local model has no per-token price. Ask it a thousand questions and the bill is your electricity. The honest trade-off is capability — local models are smaller than frontier cloud models, so keep the hard reasoning tasks for a cloud key and let the local model carry everyday work.
Ollama: the quickest route
Install Ollama, pull a model with a single command (for example a Llama, Qwen or Gemma build sized for your RAM), and it serves the model locally. In More AI, open Settings → Providers and enable Ollama — no API key, no configuration. The app picks up your installed models automatically, and they appear in the model picker next to your cloud models.
From there it behaves like any other provider: pick the local model for a conversation and start typing. The context-window meter and per-conversation model switching all work the same.
LM Studio: a friendly model manager
Prefer a graphical app? LM Studio lets you browse and download open models, then serves them from a local server with one click. More AI supports it as a first-class provider, the same way as Ollama: enable it in Settings → Providers and your downloaded models show up in the picker.
Any other server: custom endpoints
Running vLLM, llama.cpp or a LiteLLM proxy? Add a custom provider and point it at any endpoint that speaks the OpenAI-compatible Chat Completions API (the base URL usually ends in /v1) — or at an Anthropic-compatible Messages endpoint. That covers practically every self-hosted model server, on your machine or on a box in your LAN.
Mix local and cloud
You don’t have to choose once. Keys and local providers live side by side, and every conversation picks its own model — so a practical setup is a local model for private notes and quick questions, and a frontier model for the tasks that deserve it. Switching is two clicks, not a reconfiguration.
All three modes work with local models — Chat, Cowork and Code all run through the same engine. For the agent modes, prefer the most capable local model your hardware can hold: planning and tool use reward a stronger model.
Frequently asked questions
Is running a local model really free?
There is no per-token cost — the model runs on your own hardware, so the only cost is the electricity your computer uses. The More AI app itself is free too.
Does everything work offline?
Chat, Cowork and Code all work fully offline with a local model — files, documents and repositories are on your disk anyway. Only features that inherently need the internet (web search, sync, connectors) wait until you’re back online.
What hardware do I need?
A small model runs comfortably in 8–16 GB of RAM; larger, more capable models want more memory or a GPU. Apple-silicon Macs are particularly good hosts. Start small, see what your machine handles, and size up.
Which local servers does More AI support?
Ollama and LM Studio are built in — no key, no setup beyond enabling them. Beyond that, any OpenAI-compatible endpoint (vLLM, llama.cpp, LiteLLM) or Anthropic-compatible endpoint works as a custom provider.
Can the coding agent use a local model?
Yes — Code mode runs through the same engine, so a local model can plan and edit too. Agent work rewards stronger models, so use the most capable local model your hardware can hold, or switch that one session to a cloud key.