Proxy · DWG 01
opencodex
A local proxy that lets Codex and Claude Code run on the model you choose
opencodex is a local provider proxy that sits in front of OpenAI Codex and Claude Code. Codex CLI, the app or the SDK sends requests to a local address, and opencodex translates them for Claude, Gemini, Grok, DeepSeek, Ollama and other models. The Codex binary is never patched.
Spec
- Install
npm install -g @bitkyc08/opencodexocx start- Category
- Proxy
- License
- MIT
- Repository
- github.com/lidge-jun/opencodex
- Docs
- opencodex.me
- GitHub · npm
- 16,000+ stars11,000+ weekly downloadsSeptember 25, 2026 as of
Features
- FIG. 01
Two-way Responses API translation
Converts the Responses API that Codex speaks into each provider's API, and turns streaming, tool calls, reasoning and images back into Codex's format.
- FIG. 02
One listener, three front doors
The same local port accepts Responses, Anthropic Messages and OpenAI Chat Completions requests, so Claude Code can use the proxy too.
- FIG. 03
Account pools and quota-aware routing
Pool several ChatGPT accounts, watch 5-hour and weekly usage in the dashboard, and send new sessions to the account with headroom.
- FIG. 04
Local dashboard
Run
ocx startand manage providers, routing and usage from localhost.
Use it when
- You like how Codex works but want to try it on another lab's model
- Your team mixes several models but wants one agent tool
- You want an offline coding agent on local models through Ollama
FAQ
Does it fork or patch Codex?
No. opencodex runs as a separate proxy process and Codex is simply pointed at a local address. Stop the proxy and Codex goes back to normal.
Which models can I use?
Claude, Gemini, Grok, DeepSeek, Ollama and many providers with OpenAI-compatible APIs. The supported list is kept current at opencodex.me.
What is the license?
MIT. The source lives at github.com/lidge-jun/opencodex.