⚡ AnyCodex

Run ANY LLM (Meta Muse Spark, DeepSeek, Groq, Ollama, OpenRouter) inside OpenAI Codex & ChatGPT Desktop with 100% Tool Parity — Full Computer Use (CUA), Code Mode, and Zero Subscription Limits.

---

🌟 Why AnyCodex?

OpenAI Codex and the official ChatGPT Desktop application are revolutionary tools for coding, pairing, and automating your screen with Computer Use. However:

• OpenAI rate limits and usage caps quickly run out during heavy programming sessions.
• Developers often want to use unlimited free models like Meta Muse Spark, ultra-fast models like Groq, cost-effective powerhouses like DeepSeek, or 100% private local offline models via Ollama.
• Until now, switching providers broke tools, corrupted thread histories, or required sacrificing desktop automation.

AnyCodex completely solves this. It functions as an ultra-fast, local translation gateway and profile manager that connects Codex Desktop to any OpenAI-compatible LLM endpoint with:

• 🖱️ 100% Full Computer Use (CUA) Parity: Real mouse clicks, keystrokes, screenshots, window control, and browser automation.
• 💻 100% Code Mode Parity: Instant local bash execution, terminal interaction, file editing, and test verification.
• 🪟 Side-by-Side Dual-App Mode: Run official ChatGPT Plus (OpenAI) and AnyCodex at the exact same time without profile or session conflicts.
• 🔄 Instant 1-Command Hot-Switching: Seamlessly toggle between providers (anycodex use meta, anycodex use deepseek, anycodex use openai).
• 🔒 100% Local & Zero Telemetry: Operates exclusively on 127.0.0.1:8765. Your API keys, code, and prompts never touch third-party servers.

---

🚀 1-Command Quick Installation

🍏 macOS (Terminal)

Run this single command in your Terminal:

curl -fsSL https://raw.githubusercontent.com/emonibnmustafa/anycodex/main/install.sh | bash

🪟 Windows PC (PowerShell)

Open PowerShell as Administrator and run:

irm https://raw.githubusercontent.com/emonibnmustafa/anycodex/main/install.ps1 | iex

What the installer does automatically:

1. Sets up the local AnyCodex gateway in ~/.codex/anycodex.
2. Starts the zero-config background daemon (launchd on macOS, Startup service on Windows).
3. Adds fast CLI shortcuts (anycodex, usemeta, useopenai).
4. Offers to create the standalone Side-by-Side App so you can run ChatGPT Plus and AnyCodex concurrently!

---

🪟 Side-by-Side Dual-App Mode (The Superpower)

Want to keep your official ChatGPT Plus app for OpenAI models, while running a dedicated second app for unlimited free Meta Muse Spark or DeepSeek?

Run:

anycodex dual-app

App	Location	Model Used	Profile & Storage	Best For
ChatGPT (Official)	Applications / Start Menu	OpenAI GPT-5.6 Sol	Default (~/.codex)	Official Plus account tasks
AnyCodex	Applications / Desktop	Meta Muse Spark / Custom	Isolated (~/.codex-anycodex)	Unlimited free heavy coding & long pairing sessions

• Zero Extra Storage: On macOS, utilizes APFS Copy-on-Write cloning (0 MB extra disk space).
• Independent Profiles: Separate Chromium user data directories and separate SQLite databases ensure windows, cookies, and chat threads NEVER collide.
• Keep Both Open: Run both apps side-by-side simultaneously on your screen!

---

🕹️ CLI Command Reference

AnyCodex comes with an intuitive, cross-platform CLI:

Command	Shortcut	Description
anycodex status	—	View current active provider, base URL, model, and gateway health
anycodex dual-app	—	Setup or refresh the standalone Side-by-Side desktop application
anycodex use meta	usemeta	Switch to Meta Muse Spark (Unlimited Free for contributors)
anycodex use <provider>	—	Switch to any configured provider (deepseek, groq, ollama, etc.)
anycodex use openai	useopenai	Revert to official ChatGPT Plus (OpenAI) native mode
anycodex set-key <prov> <key>	setmeta <key>	Update the API key for a provider instantly
anycodex add	—	Interactive prompt to register any custom base URL & model
anycodex list	—	List all configured providers
anycodex restart	—	Restart the local background adapter gateway

---

🌐 Supported Providers Out of the Box

1. Meta AI (Muse Spark)

• Model: muse-spark-1.3-contributor
• Key Feature: Unlimited free usage with full CUA desktop automation.
• Setup:

  anycodex set-key meta "YOUR_META_KEY"
  anycodex use meta

2. DeepSeek (DeepSeek Coder / V3)

• Model: deepseek-coder
• Base URL: https://api.deepseek.com
• Setup:

  anycodex set-key deepseek "sk-YOUR_DEEPSEEK_KEY"
  anycodex use deepseek

3. Groq (Ultra-Low Latency)

• Model: llama-3.3-70b-versatile
• Base URL: https://api.groq.com/openai
• Setup:

  anycodex set-key groq "gsk_YOUR_GROQ_KEY"
  anycodex use groq

4. Local Ollama (100% Offline & Private)

• Model: qwen2.5-coder:latest (or any installed model)
• Base URL: http://127.0.0.1:11434
• Setup:

  ollama run qwen2.5-coder
  anycodex use ollama

5. OpenRouter / Custom Endpoints

• Connect any OpenAI-compatible API (vLLM, LMStudio, Mistral, Together AI, Claude via OpenRouter):

  anycodex add

---

🏗️ How It Works (Architecture)

┌────────────────────────────────────────────────────────┐
│               ChatGPT Desktop / Codex App              │
│       Code Mode (exec) + Computer Use (CUA) Tools      │
└──────────────────────────┬─────────────────────────────┘
                           │ HTTP / SSE Stream
                           ▼
┌────────────────────────────────────────────────────────┐
│           AnyCodex Gateway (127.0.0.1:8765)        │
│  • Bidirectional Tool Translation (exec <-> CUA)       │
│  • Strict-Mode Schema Flattening & Normalization       │
│  • 64-Character Tool Name Length Clamping & Mapping    │
│  • Real-time SSE Stream Event Translation              │
└──────────────────────────┬─────────────────────────────┘
                           │ Standard OpenAI-Compatible API
                           ▼
┌────────────────────────────────────────────────────────┐
│                  Upstream Provider                     │
│    Meta AI | DeepSeek | Groq | Ollama | OpenRouter     │
└────────────────────────────────────────────────────────┘

Key Technical Innovations

1. Bidirectional Code Mode Bridge: Translates Codex's internal V8 isolate custom_tool_call on exec into standard OpenAPI function calls, and maps upstream results back to Codex flawlessly.
2. Strict-Mode Schema Sanitizer: Upstream function-calling validators (like Meta's) reject combiners like oneOf/anyOf/allOf at the root parameters level. AnyCodex flattens and normalizes schemas into clean objects.
3. Deterministic 64-Character Name Mapper: Codex plugins often emit 70–90 character tool names. AnyCodex safely clamps and bi-directionally maps names under the 64-character limit.

---

🗑️ Uninstallation

If you ever wish to completely remove AnyCodex:

macOS:

curl -fsSL https://raw.githubusercontent.com/emonibnmustafa/anycodex/main/uninstall.sh | bash

Windows:

irm https://raw.githubusercontent.com/emonibnmustafa/anycodex/main/uninstall.ps1 | iex

---

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request or open an Issue for new provider presets, feature requests, or bug reports.

📄 License

This project is licensed under the MIT License.