# Local models & offline work

Connect a model service you manage and understand the local workflow.

Updated: 2026-10-05
Desktop 0.5.0-rc.35 · Web 0.5.0-rc.36
Canonical: https://ai.z-group.cn/docs/en/local-models/

## Does MATLOOK Agent include a local model?
No. MATLOOK Agent does not bundle or automatically start a model runtime. Prepare Ollama, LM Studio or another compatible service yourself, then add a connection in MATLOOK Agent.
## Configuration order
1. Install, configure and start the service on the execution computer, and load a model.
2. Confirm the model ID, endpoint and authentication requirements in that service.
3. In MATLOOK Agent Settings → AI, choose an appropriate compatible connection and enter those details.
4. Test plain text first, then the tool or image capabilities your task requires.
5. After installing all dependencies, disconnect the network and verify the workflow you actually need.
## Offline boundaries
| Work | Requirement |
|---|---|
| Local knowledge editing and full-text search | Local sources and index are available |
| Installed tools and skills | Scripts, templates and dependencies are installed; the action itself does not need a network |
| Model inference | Local service is running and supports the task |
| Online models, web collection and remote sync | Network access and the relevant service |
## Common confusion
`localhost` and `127.0.0.1` refer to the computer running the connection. Using that address on ECS does not reach a model in your office. A model that chats successfully may still lack suitable tool-calling or vision support.