buddylist

Product Introduction

1. Definition: buddylist is a free, cross-platform desktop application (macOS Apple Silicon, Windows, Linux) that functions as a unified, retro-styled interface for managing multiple AI coding agents. It is a technical productivity tool that integrates directly with the command-line interfaces (CLIs) of major AI coding assistants—Claude Code, Codex, Gemini, and OpenCode—presenting them as contacts in an MSN Messenger-style buddy list.
2. Core Value Proposition: buddylist exists to solve the workflow fragmentation of using multiple AI coding agents. It centralizes management, enabling developers to orchestrate collaborative work between different AI models on the same codebase, track progress visually, and maintain a clean, isolated git worktree for each agent, all within a single, familiar chat interface.

Main Features

1. Multi-Agent Buddy List Interface: The core interface is a contact list where each "buddy" represents an AI agent connected to a specific project folder. Contacts display real-time status (Online, Busy, Away), pull request check status (e.g., "2 failing", "Passed"), and are grouped by project, service, or custom rules. Double-clicking a contact opens a dedicated chat session with that agent in its assigned project directory.
2. Git Worktree & Branch Isolation: For any project under git version control, buddylist automatically creates a separate git worktree and branch for each agent contact. This is a critical technical feature that prevents file conflicts by ensuring each AI operates on an isolated copy of the codebase, while keeping the main project folder untouched.
3. Collaborative Multi-Agent Group Chats: Users can invite multiple AI agents (e.g., Claude, Codex, Gemini) into a single conversation thread. Agents can be configured to answer in sequence ("Take turns") or simultaneously ("All at once"). They can hand off tasks to each other using @mentions, and an "effort check" feature allows the agents themselves to propose the appropriate thinking level (Low, Medium, High, Max) for each participant before responding.
4. Integrated Development Context Pane: A dedicated sidebar provides tabs for real-time development context alongside the chat. This includes a Plan tab for reviewing and approving AI-generated plans, a Changes tab displaying a file tree and granular diffs with inline comment threads, a Browser tab for previewing a local dev server, a Subagents tab for monitoring Claude's sub-tasks, a Session info panel, and system Terminals.
5. Agent-Action Management & Permissions: Instead of automatic execution, AI agents send "invitations" for actions like running shell commands or editing files. The user must Accept, "Always allow," or Decline each request. All tool calls and their outputs (commands, file edits, screenshots) are logged as discrete, inspectable lines within the chat history.

Problems Solved

1. Pain Point: Context switching and terminal sprawl when juggling multiple AI coding assistants (Claude, Codex, etc.) across different projects.
2. Target Audience: Software developers, engineering teams, and technical leads who regularly use AI pair programmers. Specifically, full-stack developers, DevOps engineers, and open-source maintainers who manage multiple feature branches and pull requests.
3. Use Cases:
   • Orchestrated Code Reviews: Assign Claude to implement a feature, then automatically @mention Codex within the same chat to review the code, and finally ask Gemini to write the tests.
   • Parallel Experimentation: Using separate contacts/agents on the same project to test different implementation approaches for a bug fix in isolated git worktrees.
   • Pull Request Monitoring: Visually tracking CI/CD check statuses (passing/failing) for multiple active branches directly from the contact list without switching to GitHub.
   • Legacy Project Onboarding: Using a group chat with multiple AI agents to collectively analyze and explain a complex, unfamiliar codebase.

Unique Advantages

1. Strengths & Limitations (Pros & Cons):
   • Pros:
     • Unified Orchestration: Unparalleled ability to manage and coordinate multiple AI agents from different providers in a coherent workflow.
     • Git Integrity: Native git worktree isolation is a best-practice implementation that protects the main branch and prevents agent interference.
     • Visual & Contextual Workflow: Integrates chat, code diffs, browser preview, and terminals into a single window, reducing tab overload.
     • Bring-Your-Own-Account (BYOA): No locked-in subscription; uses your existing Anthropic, OpenAI, Google, or OpenCode API accounts and CLI configurations.
   • Cons:
     • Learning Curve: The MSN Messenger metaphor and dense feature set require initial learning compared to a simple terminal or web chat.
     • Desktop Application: Requires download and installation, unlike purely web-based tools. Updates must be managed.
     • Provider Dependency: Functionality for Codex, Gemini, and OpenCode depends on the user correctly installing and configuring their respective CLIs.
2. Key Alternatives & Differentiation:
   • Individual Provider IDEs/CLIs (Claude Dev, Cursor, Windsurf): These are single-agent environments deeply integrated into an editor. buddylist differentiates by being agent-agnostic, focusing on orchestration between agents rather than deep editing within one.
   • Terminal Tabs with Multiple CLI Sessions: The manual alternative. buddylist provides a superior visual management layer, status overview, structured chat history, built-in git worktree management, and collaborative features that plain terminals lack.
   • AI Agent Orchestration Platforms (e.g., GPT Engineer, Smol Agents): These are typically code-first frameworks for automating multi-step AI workflows. buddylist differentiates by being a real-time, interactive, and visual desktop GUI designed for human-in-the-loop collaboration, not fully automated pipelines.

Frequently Asked Questions (FAQ)

1. Is buddylist a chatbot or a code editor? buddylist is neither a chatbot UI skin nor a full code editor. It is an orchestration layer and visual interface that connects to real AI coding agent CLIs. You edit code in your preferred external editor (opened via buddylist's "Open" button), while buddylist manages the conversation, permissions, file changes, and agent coordination.
2. How does buddylist handle API costs and usage? buddylist operates on a Bring-Your-Own-Account model. It uses the API keys and configurations from your locally installed Claude Code, Codex, Gemini, and OpenCode CLIs. All usage costs are billed directly to you by the respective AI provider (Anthropic, OpenAI, Google); buddylist itself is a free application and does not charge for API access.
3. Can I use buddylist with GitHub Copilot? Not directly. buddylist integrates with AI agents that offer a command-line interface (CLI) for programmatic interaction, such as Claude Code (claude), Codex (codex), and Gemini (gemini). GitHub Copilot is primarily an IDE extension and does not provide a standalone CLI for this type of external orchestration.
4. How does the git worktree feature work? When you add a project under git version control as a contact, buddylist uses the git worktree add command to create a separate, linked working directory for that agent. Each agent's changes are committed to a unique branch. This keeps the main project directory clean and allows you to instantly switch contexts between agents without stashing or committing unfinished work.
5. What are the system requirements for buddylist? buddylist is a native desktop application available for macOS (Apple Silicon), Windows, and Linux. It requires the respective AI provider's CLI tools (e.g., Anthropic's Claude CLI) to be installed and authenticated on your system. Specific RAM or CPU requirements are not listed but scale with the underlying AI model usage.