codefa

The High-Performance Local AI Proxy & Control Center for Coding Agents

---

🚀 Overview

codefa is an ultra-fast, local-first proxy and management suite designed to connect autonomous AI coding agents—such as Claude Code and OpenAI Codex—to any OpenAI-compatible AI provider (NVIDIA NIM, DeepSeek, OpenRouter, Mistral, Ollama, LM Studio, Kimi, MiniMax, and more).

Featuring an interactive terminal launcher, a modern responsive Admin UI on port 8090, and live HTTP access log streaming, codefa gives you total control over model routing, provider API keys, and server diagnostics.

Interactive borderless CLI selection menu in terminal

---

✨ Features

• ⚡ Interactive Terminal Launcher (codefa): Arrow-key navigation (↑/↓) to seamlessly launch Claude Code, Codex CLI, or Server Only.
• 🛠️ Responsive Admin Control Panel: Beautiful glassmorphism Web UI running on http://127.0.0.1:8090/admin with full Dark & Light mode support.
• 📜 Live Server Logs Stream: Complete real-time capture of Uvicorn HTTP access logs, API payloads, and internal routing events.
• 🤖 Universal Coding Agent Compatibility: Pre-configured support for both Anthropic Claude Code and OpenAI Codex CLI.
• 🔒 100% Local & Secure: API keys remain strictly stored in your local ~/.codefa/.env file with zero telemetry leakage.
• 🏎️ Optimized Performance: Built on Python 3.14, FastAPI, Loguru, and uvloop for minimal resource overhead.

---

📦 Installation

1. Install Or Update

macOS / Linux

curl -fsSL https://codefa.foshati.com/install.sh | bash

Windows PowerShell

irm https://codefa.foshati.com/install.ps1 | iex

2. Start The Server

codefa server

---

⚡ Quick Start

1. Launching codefa

Run codefa in your terminal to open the interactive selection menu:

codefa

Use the ↑ and ↓ arrow keys to highlight your choice, then press Enter:

• Claude Code: Launches Anthropic Claude Code CLI configured with codefa proxy.
• Codex CLI: Launches OpenAI Codex CLI configured with codefa proxy.
• Server Only: Starts the background proxy server and opens the Admin UI on port 8090.

Direct Subcommands

You can also bypass the menu and launch directly:

codefa claude   # Direct launcher for Claude Code
codefa codex    # Direct launcher for Codex CLI
codefa server   # Start server & open Admin UI
codefa-init     # Initialize ~/.codefa/.env config

---

## Choose A Provider

Provider	Description	Example Model
NVIDIA NIM	Accelerated NIM microservices	nvidia_nim/meta/llama-3.3-70b-instruct
OpenRouter	Unified multi-provider gateway	open_router/auto
Gemini	Google Gemini API	gemini/gemini-2.5-flash
Vertex AI	Google Cloud Vertex AI	vertex/gemini-2.5-flash
DeepSeek	DeepSeek V3 & R1 models	deepseek/deepseek-chat
Mistral	Mistral AI API endpoints	mistral/mistral-large-latest
Mistral Codestral	Codestral IDE endpoint	mistral_codestral/codestral-latest
OpenCode Zen	OpenCode Zen API	opencode/claude-3-5-sonnet
OpenCode Go	OpenCode Go API	opencode_go/claude-3-5-sonnet
Vercel AI Gateway	Vercel AI Gateway	vercel/openai/gpt-4o
Amazon Bedrock	Amazon Bedrock Mantle	bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0
Hugging Face	Hugging Face Router	huggingface/meta-llama/Llama-3.3-70B-Instruct
Cohere	Cohere API	cohere/command-r-plus
GitHub Models	GitHub Models endpoint	github_models/gpt-4o
Wafer	Wafer AI Pass	wafer/claude-3-5-sonnet
Kimi	Moonshot Kimi API	kimi/moonshot-v1-8k
Kimi Code	Kimi Code Subscription	kimi_code/kimi-k1.5
MiniMax	MiniMax AI API	minimax/MiniMax-Text-01
Cerebras	Cerebras Inference Engine	cerebras/llama3.1-70b
Groq	Groq LPU Inference	groq/llama-3.3-70b-versatile
SambaNova	SambaNova Cloud	sambanova/Meta-Llama-3.3-70B-Instruct
Fireworks	Fireworks AI	fireworks/accounts/fireworks/models/llama-v3p3-70b-instruct
Cloudflare	Workers AI catalog	cloudflare/@cf/meta/llama-3.3-70b-instruct
Z.ai	Z.ai GLM Coding Plan	zai/glm-4
Ollama Cloud	Ollama Cloud remote	ollama_cloud/llama3.3
LM Studio	Local LM Studio server	lmstudio/local-model
llama.cpp	Local llama.cpp server	llamacpp/local-model
Ollama	Local Ollama instance	ollama/llama3.3:latest

---

🎨 Admin UI & Control Center

Access the Admin UI anytime at http://127.0.0.1:8090/admin.

Providers Configuration

Configure API keys, model choices, rate limits, and proxies for all supported AI providers in a responsive, theme-aware Web UI:

Live Server Logs

Monitor live HTTP requests, status codes, and routing events in real time:

---

🤖 Coding Agents Integration

Claude Code

codefa automatically injects proxy environment variables for seamless integration with Anthropic Claude Code:

OpenAI Codex CLI

Run Codex CLI with custom models and local provider routing:

---

🏗️ Architecture & Workflow

codefa acts as a local bridge between client protocol requests (Anthropic Messages API & OpenAI Responses API) and downstream AI provider endpoints:

---

🗑️ Uninstallation

To uninstall codefa:

macOS / Linux

curl -fsSL "https://raw.githubusercontent.com/Foshati/codefa/main/scripts/uninstall.sh" | sh

Windows PowerShell

& ([scriptblock]::Create((irm "https://raw.githubusercontent.com/Foshati/codefa/main/scripts/uninstall.ps1")))

This verifies every CODEFA command is gone.

---

📄 License

Distributed under the MIT License. See LICENSE for more information.

Created with ❤️ by Foshati