nana

Product Introduction

1. Definition: Nana is a native terminal multiplexer application for Windows, built with Rust, GPUI, and the libghostty terminal engine. It is specifically designed to manage multiple AI-powered coding agents (like Claude Code, Cursor, and OpenCode) concurrently within a single, organized workspace.
2. Core Value Proposition: Nana exists to solve the organizational chaos of running multiple AI coding assistants simultaneously. Its core value is enabling developers to run every agent in isolated Git worktrees within a fast, native terminal interface, providing clear agent status, integrated code review, and persistent background sessions without requiring administrative privileges.

Main Features

1. Git Worktree Integration: Each AI coding agent operates within its own dedicated Git worktree. This technical approach forks the repository's state, allowing agents to work on separate branches (e.g., feat/auth, fix/race-pty) completely independently. Users can switch, merge, and clean up these worktrees directly from Nana's sidebar, preventing file conflicts and maintaining a clean git history.
2. Agent Status Monitoring: Nana automatically detects agents running in its terminal panes and displays their real-time status: Working, Waiting, or Done. This functionality requires no additional hooks or configuration, providing immediate visibility into which agent is active, needs input, or has completed a task, streamlining the developer's oversight workflow.
3. Persistent Background Sessions (nana-server): Terminal sessions are managed by a separate background server process (nana-server). This architecture allows users to close the main Nana application window while their AI agents continue running. Sessions persist between app launches, and the feature also supports deploying nana-server on remote Linux/macOS machines for SSH-based development.
4. Built-in Code Review Tools: Nana includes integrated tools for reviewing code changes directly within the terminal environment. Users can view Git diffs of the current changes, browse commit history, and revisit previous modifications without switching to a separate Git GUI or IDE, keeping the workflow contained.
5. Composer for Multi-Pane Prompts: A "Composer" feature allows users to craft complex prompts, potentially including file references and images, and then broadcast them to a single terminal pane or to every split pane simultaneously. This is designed for efficiently issuing the same instruction to multiple AI agents at once.

Problems Solved

1. Pain Point: Managing multiple AI coding agents leads to terminal tab sprawl, workspace confusion, and agents overwriting each other's work. Manually tracking which agent is doing what and reviewing their disparate outputs is inefficient.
2. Target Audience: The primary user personas are Windows-based software engineers, developers, and technical leads who actively use multiple AI coding assistants (Claude Code, Cursor, etc.) in their daily workflow and require a structured, production-oriented approach to parallelize and supervise AI-generated code.
3. Use Cases:
   • Parallel Feature Development: Assigning different AI agents to work on separate features or bugs in isolated Git worktrees simultaneously.
   • Code Review & Testing Supervision: Using the status monitor and diff viewer to oversee agent progress, provide prompts when they are "waiting," and review code changes before integration.
   • Long-Running Agent Sessions: Starting a complex code generation or refactoring task with an agent, closing the laptop, and reconnecting later to check results via the persistent nana-server.
   • Unified Project Management: Consolidating all terminal-based AI agent activity for a single project into one organized, themeable native application instead of scattered terminal windows and IDE plugins.

Unique Advantages

1. Strengths & Limitations (Pros & Cons):

   • Pros:
     • Native Performance: Built with Rust and GPU-accelerated UI (GPUI), resulting in a fast, memory-efficient application unlike Electron-based tools.
     • Core Workflow Innovation: Deep integration of Git worktrees with terminal multiplexing is a unique and powerful solution for multi-agent AI development.
     • Zero-Config Agent Detection: Automatic status reporting for supported agents improves usability without setup overhead.
     • Truly Persistent Sessions: The client-server model ensures work continues after the UI is closed, a significant advantage for long tasks.
     • Free Beta: Currently free to use with no announced pricing, lowering the barrier to entry.
   • Cons:
     • Platform Lock-in: Currently exclusive to Windows 10/11, excluding macOS and Linux desktop users (though remote servers are supported).
     • Agent Ecosystem Dependency: Its value is tied to compatibility with specific AI agents (Claude Code, Cursor, etc.). Support for new or custom agents may vary.
     • Learning Curve: The paradigm of worktree-per-agent and the multiplexer interface may require adaptation for users accustomed to simpler terminal workflows.
     • Beta Status: As a beta product, long-term stability, feature set, and future pricing are subject to change.

2. Key Alternatives & Differentiation:

   • Traditional Terminal Multiplexers (tmux, screen): While powerful, they lack native AI agent awareness, integrated Git worktree management, and a dedicated GUI. Nana provides a higher-level, opinionated workflow specifically for AI coding, whereas tmux is a general-purpose, configuration-heavy tool.
   • IDE-Based AI Tools (Cursor, VS Code with Extensions): These integrate AI directly into the editor. Nana differentiates by being agent-agnostic and terminal-centric, allowing use of any CLI-based agent (including those from IDEs run in headless mode) and focusing on parallel execution and project-level organization across multiple agents, rather than deep editing features.
   • Windows Terminal: This is a general-purpose terminal emulator. Nana is a layer on top, providing the specialized multiplexing, session persistence, and agent management features that Windows Terminal does not offer, though it may utilize similar underlying rendering tech (libghostty).

Frequently Asked Questions (FAQ)

1. Is Nana a replacement for Windows Terminal or my IDE? No, Nana is a complementary tool. It is a specialized terminal multiplexer that organizes AI agent workflows. You would still use it in conjunction with Windows Terminal for other tasks and an IDE for detailed code editing.
2. How does Nana handle different AI agents and their configurations? Nana acts as a management layer for the terminal sessions in which agents run. It detects common agents by their process/output. You configure and run the agents themselves (e.g., Claude Code, OpenCode) within Nana's panes as you normally would in any terminal, and Nana adds the organizational features on top.
3. What are the system requirements for running Nana? Nana requires Windows 10 or newer. It is a native application built with Rust and does not require administrator rights to install or run. It leverages WebView2 for its extension system, which is a standard Windows component.
4. Can I use Nana for non-AI development work? Yes, while optimized for AI agents, its core functions—terminal multiplexing, Git worktree management, persistent sessions, and themeing—are beneficial for any multi-terminal development workflow on Windows, such as running servers, logs, and build processes side-by-side.
5. How does the pricing work for the Nana beta? Nana is currently in a free beta period. The official website indicates a "Free beta" with no immediate cost. Future pricing plans have not been detailed, so users should monitor official channels for updates on a potential transition to a paid model.