Models
LM Studio and Apple Foundation
Two more local engines 8gent can use with no key.
LM Studio
Start LM Studio's local server (default port 1234) with a model loaded. 8gent checks for it on first run and lists it as a provider.
/provider lmstudio
/model <model id>| Variable | What it does | Default |
|---|---|---|
LMSTUDIO_BASE_URL | Where 8gent sends LM Studio requests | http://localhost:1234/v1 |
LM_STUDIO_HOST | Where the first-run check looks for LM Studio | port 1234 on this machine |
By default the Engineer tab is set to use LM Studio (models.tabs.engineer in ~/.8gent/settings.json). If LM Studio is not running, that tab falls back through the failover chain.
Apple Foundation
On a Mac with macOS 26 or later on Apple Silicon, 8gent can use the on-device Apple Foundation model through a small local bridge, apfel, on port 11435.
- The provider switches itself on when the Mac qualifies, and stays off elsewhere.
- From a source checkout,
bun run install:apple-foundationinstalls the bridge. APFEL_BASE_URLpoints 8gent at a bridge on another address.
The on-device model does not use tools or see images, so it suits short answers and quick checks rather than editing code. By default it is set for the QA tab.