Codexia is an ambitious new open-source project attempting to create a proper IDE experience for developers running Anthropic's Claude Code and OpenAI's Codex CLI agents side-by-side. Built with Tauri v2 and Rust, the app wraps both agent runtimes in a single desktop workspace that includes file management, prompt drafting tools, task scheduling, and even a headless web server for remote control.

Core Features

The application isn't just a fancy terminal wrapperβ€”it brings actual IDE-level functionality to the agent workflow. Users get Git worktree management built directly into the sidebar, an IDE-like editor with project file tree navigation, a dedicated prompt notepad for drafting instructions before launching sessions, and local web preview capabilities for testing frontends without leaving the app. The Task Scheduler feature handles recurring automation jobs, which is something neither Codex nor Claude Code natively support out of the box.

Architecture Deep Dive

Under the hood, Codexia uses React + TypeScript with Zustand state management and shadcn/ui components on the frontend. The desktop backend runs Tauri v2 with Rust handling native integrations while an Axum-based web server exposes HTTP endpoints for headless operation. When running in headless mode, clients can connect via WebSocket at /ws for real-time streaming updates. The agent runtime itself integrates with Codex's app-server JSON-RPC protocol for session and turn lifecycle managementβ€”a non-trivial implementation detail that suggests serious engineering effort went into this project.

Security Model

Codexia takes a notably conservative approach to security given its nature as an AI agent wrapper. Agents execute in isolated processes, file and network permissions are configurable per-agent, all data remains local by default, and telemetry is opt-in only. The AGPL-3.0 license ensures open-source transparency for the core project while a separate commercial license allows closed-source SaaS deploymentsβ€”a dual-licensing setup that lets enterprise users avoid viral GPL obligations.

Getting Started

Installation options include Homebrew for macOS users via brew tap milisp/codexia followed by brew install --cask codexia, or prebuilt binaries for macOS, Linux, and Windows from GitHub Releases. Users need both Codex CLI and Claude Code CLI installed beforehand as explicit requirements. The quick start guide walks through adding a project directory, entering prompts, starting agent sessions, and creating scheduled automation jobs in under five minutes.

Community Ecosystem

The project has already attracted several interesting forks worth watching: jeremiahodom/codex-ui swaps the Rust backend for Node.js with API/SSE support, Itexoft/codexia adds SSH integration capabilities, and nuno5645/codexia contributes reasoning and token count event tracking. Related projects include awesome-codex-cli as a curated resource list and claw-army/claude-node as a Python subprocess bridge specifically for Claude Code CLI.

The Bottom Line

Codexia represents the kind of infrastructure layer the AI agent ecosystem desperately needsβ€”bridging multiple competing runtimes under one roof while adding workflow features the CLIs themselves lack. Whether it gains traction depends heavily on how quickly Anthropic and OpenAI solidify their respective agent APIs, but for developers already running both systems, this looks like a genuinely useful convergence point worth watching.