[Official FAQ] Corrections and follow-up questions welcome.

Yes. Continue can call a local OpenAI-compatible endpoint (Ollama, llama.cpp server, vLLM, etc.) for chat and autocomplete so source code stays on your machine.

Setup ideas:

• Run a coding-oriented model first and verify chat works outside the IDE.
• Use a smaller/faster model for inline completion and a larger one for chat/refactors if VRAM allows.
• Large IDE context windows increase latency and memory use—start moderate.

Local models may lag frontier cloud models on hard refactors; measure on your repo. Approximate sizing only—benchmark on your hardware.