From the Codex × Ollama video

Get the Codex × Local Cheat Sheet

OpenAI's official Codex desktop app has an open provider layer — point it at Ollama on localhost and the entire Codex agent UI runs against any open-weight model on your machine. Zero API bills, zero rate limits, zero data leaving your laptop. This sheet gives you the exact 8-line ~/.codex/config.toml block, the 8-step Ollama setup, a model-by-VRAM picker (Gemma 4, Qwen, Llama, Deepseek), the FAQ that covers the gotchas, and the four reasons this matters more than OpenAI is letting on.

Join 25,000+ readers of the Hyperautomation AI ReportGet the PDF freeKeyword: CODEX

By submitting, you'll be subscribed to the Hyperautomation AI Report newsletter — three practical AI emails a week. Unsubscribe anytime.

Free. No spam. Unsubscribe anytime.