Run NJORD Engine on local models with Ollama and LM Studio
You can now point the AI agents at a local model running in Ollama or LM Studio, with no cloud key required, and nothing leaves your machine.
NJORD Engine is bring-your-own-keys (BYOK), and from this build that includes bringing your own model. Alongside cloud providers, you can now run the agents against a local model served by Ollama or LM Studio.
Why local models
- Privacy. Your prompts and project never leave your computer.
- No per-token cost. Once the model is on your disk, generating is free.
- Offline. Keep building on a train, a plane, or a cabin with bad wifi.
How to set it up
- Install Ollama or LM Studio and pull a model (a capable coding model is recommended for the build agents).
- Start the local server.
- In NJORD Engine, open Settings â–¸ AI providers, set a capability to Local / OpenAI-compatible, and point it at the local server address.
That's it. The same agents that build your game now run on your own hardware.
Cloud models still tend to be stronger for the heaviest reasoning, so you can mix and match: a local model for quick iteration, a cloud key for the big lifts. See the Getting started guide for the full setup.