Problem solved

Coding assistants like OpenAI Codex and Claude Code typically require using their own proprietary models, making it difficult to switch to alternative LLMs such as Gemini, Grok, DeepSeek, or local models via Ollama. Opencodex solves this by providing a universal proxy that lets these tools route requests to any supported LLM provider, reducing vendor lock-in and enabling flexible model selection.

Use cases

• Use Gemini models with OpenAI Codex CLI
• Run Claude Code with local Ollama models
• Connect Codex SDK to DeepSeek API
• Switch Codex App between multiple cloud LLM providers
• Self-host a proxy to centralize LLM access for development teams
• Test coding agent behavior across different LLM backends

Audience

• Developers
• ML Engineers
• Hobbyists

Pros

• Open source and self-hostable
• Supports multiple LLM providers including Claude, Gemini, Grok, DeepSeek, and Ollama
• Works with Codex CLI, App, SDK, and Claude Code
• Helps avoid vendor lock-in
• Enables use of local models for privacy

Cons

• Requires configuration and technical setup
• Exact provider compatibility may vary
• Limited public documentation in the provided metadata
• Depends on upstream Codex and Claude Code interfaces
• No formal support or SLA mentioned