AIRGAP StudioProvider Configuration
AI provider setup and model configuration guide
AIRGAP Assistant (Cline fork) Provider Configuration
AIRGAP Assistant supports 4 provider types that conform to the OpenAI-compatible API standard.
Default Configuration — llama-server (Automatic)
Right after installation, AIRGAP Assistant automatically uses the following endpoint:
| Field | Value |
|---|---|
| Provider | openai-compatible |
| Base URL | http://127.0.0.1:11433/v1 (via Compatibility Proxy) |
| Model | Currently active model (determined by AIRGAP Monitor) |
| API Key | Not required (local) |
The dev-launcher or installer automatically patches the Base URL to
11433(Proxy). Do not point it directly at11434(llama-server) — that conflicts with the automatic patch.
Supported Provider Types
| Provider Name | Purpose | API Key |
|---|---|---|
openai | Generic OpenAI-compatible endpoints (LM Studio, vLLM, in-house OpenAI gateway, etc.) | Required (depends on endpoint) |
openai-native | OpenAI direct (api.openai.com) | Required |
openai-codex | Codex CLI compatible | Required |
lmstudio | LM Studio compatible | Not required |
Anthropic Claude and Google Gemini are not direct providers of AIRGAP Assistant. If you need external SaaS providers, use the foundation provider opt-in on AIRGAP Lite Assistant.
Changing Providers
- AIRGAP Assistant side panel → Settings (gear icon)
- Select one of the 4 types above from the Provider dropdown
- Enter Base URL / API Key if needed
- Save → applies from the next message
Model Changes Are Separate
AI model changes are performed via the AIRGAP: Select LLM Model command in AIRGAP Monitor, not in provider settings. The provider stays put and only the model is swapped.
Details: Monitor documentation
External Network Usage
We do not recommend using external SaaS providers in air-gapped environments. Connecting to an external provider requires internet access and may be blocked by environment policies.