local-codex-proxy: run self-hosted models inside Codex alongside ChatGPT
TheZachMueller · x · 2026-10-04
TheZachMueller released local-codex-proxy, a single dependency-free Python file that lets Codex use self-hosted models while keeping ChatGPT models in the same session.
- Codex supports one model provider per session, so ChatGPT-backed sessions can't reach your own model server
- The proxy listens on 127.0.0.1:4100 and routes by model name: everything else goes to chatgpt.com's backend with auth forwarded, custom models go to your own host with auth stripped
- Translates Codex's Responses Lite nested-namespace tools into SGLang's flat function-tool form and rewrites tool calls back into namespaced form
- Strips Codex's JavaScript code-mode wrappers that small models can't reliably produce
Pair it with per-machine ChatGPT config changes; the author even points his LLM at the repo to configure itself.
Related event: Open-Source local-codex-proxy Brings Local Models Into ChatGPT(2 posts)→
More from coding & agent
- Meta's RankEvolve doubles down on agent reliability, lifting auto-research accuracy 45.8% to 62.5% — dair_ai · 2026-10-04
- CMU paper: RL-trained 4B proposer edits agent harness code, beats its 35B teacher — omarsar0 · 2026-10-04
- Dev praises pi-durable for running long-lived agents behind any interface — tokumin · 2026-10-04
- Claude Code 2.1.289 prompt tokens jump 17.8%, system prompts now over half — ClaudeCodeLog · 2026-10-04
- Claude Code 2.1.289 patches @-mention symlink bypass of file read deny rules — ClaudeCodeLog · 2026-10-04
- Claude Code 2.1.289 ships agent.spawn and fixes @-mention permission bypass — ClaudeCodeLog · 2026-10-04