An Emacs package, ai-code-interface.el, provides a unified interface for interacting with various AI coding assistants like OpenAI Codex, GitHub Copilot CLI, and Claude Code. It targets Emacs users who leverage multiple AI tools, offering a consistent user experience and workflow portability, allowing seamless switching between backends without altering core Emacs commands or interaction patterns.

How It Works

This package acts as a meta-interface, abstracting the complexities of different AI coding CLIs and external backend packages. It focuses on providing a stable Emacs workflow by standardizing context gathering, prompt formatting, and command execution across diverse AI services. Its core advantage lies in workflow portability, ensuring that user-defined actions and interactions remain consistent regardless of the underlying AI model or CLI being used.

Quick Start & Requirements

Installation is typically done via MELPA. Add MELPA to your package-archives in init.el, then run M-x package-install ai-code. Configure using use-package and set a default backend, e.g., (ai-code-set-backend 'codex). The primary interaction point is the C-c a keybinding, which opens a contextual transient menu.

Required dependencies include Emacs 29.1 or later, org, magit, transient, and a terminal emulator like vterm (default), eat, or ghostel.

Highlighted Details

• Transient-Driven Hub: A single keybinding (C-c a) accesses a comprehensive, contextual menu for all AI coding actions, eliminating the need to memorize scattered keybindings.
• Context-Aware Code Actions: Automatically captures relevant code context (current function, region, visible files, clipboard) to ensure precise AI prompts.
• Agile Development Workflows: Integrates features for refactoring, Test-Driven Development (TDD) cycles, and pull request reviews.
• MCP (Model Communication Protocol) Tools: Includes a built-in Emacs MCP server for debugging Emacs Lisp issues, inspecting editor state, and enabling AI-driven fixes via eval_elisp.
• Harness Engineering: Implements automated feedback loops for code changes, including automatic testing and follow-up actions, to create a more robust AI development process.
• Portable Agent Handoff: Utilizes Org-mode task files to facilitate transferring development context and progress between different AI coding backends.

Maintenance & Community

The project actively tracks GitHub contributors. Community interaction and support are primarily managed through GitHub issues and pull requests.

Licensing & Compatibility

The project is licensed under the Apache-2.0 License, permitting commercial use and modification with attribution. It integrates with the broader Emacs Lisp ecosystem.

Limitations & Caveats

The ghostel terminal backend may truncate long conversations due to its default scrollback limit, requiring manual configuration. Certain backends, like ECA and agent-shell, do not support the ai-code-apply-prompt-on-current-file command. Desktop notifications are experimental and may appear in the minibuffer on non-Linux platforms.