Summary

This project provides an Agent Client Protocol (ACP) adapter for the Codex CLI, enabling AI-powered code assistance within ACP-compatible clients like Zed. It targets developers seeking to integrate Codex's capabilities, such as code review, editing, and AI tool usage, directly into their development workflows via a standardized protocol. The adapter streamlines interaction with Codex, offering advanced features through familiar client interfaces.

How It Works

The core of the project is an ACP adapter that wraps the Codex CLI, translating ACP messages into Codex commands. It supports features like context-aware @-mentions, image handling, tool calls requiring explicit permission, and various slash commands for code review and management. This approach allows ACP clients to leverage Codex's underlying models and functionalities without needing direct CLI integration for each operation.

Quick Start & Requirements

• Installation: Download the latest release for your OS/architecture from the releases page or use npx @zed-industries/codex-acp.
• Prerequisites: Requires either CODEX_API_KEY or OPENAI_API_KEY environment variables. A paid ChatGPT subscription is supported but has limitations.
• Links: Releases: https://github.com/zed-industries/codex-acp/releases. Zed Integration Docs: "External Agent support".

Highlighted Details

• Implements advanced ACP features including context @-mentions, image support, and permission-based tool calls.
• Offers comprehensive slash command support for code review (/review, /review-branch, /review-commit), initialization (/init), and session management (/compact, /logout).
• Supports multiple authentication methods: CODEX_API_KEY, OPENAI_API_KEY, and a paid ChatGPT subscription.
• Designed for seamless integration with Zed and other ACP-compatible clients.

Maintenance & Community

No specific details regarding maintainers, community channels (like Discord/Slack), or project roadmap were provided in the README snippet.

Licensing & Compatibility

• License: Apache-2.0.
• Compatibility: The Apache-2.0 license is permissive and generally suitable for commercial use and integration into closed-source projects.

Limitations & Caveats

The authentication method using a paid ChatGPT subscription is explicitly noted as not functional in remote projects. The project's description in the README is minimal, indicating that detailed usage and advanced features might be primarily documented elsewhere or inferred from the feature list.