J3n5en/EnsoAI

Multiple Agents, Parallel Flow

EnsoAI – Parallel AI‑Agent Workspaces for Git Worktrees

What it is – EnsoAI is a desktop application (Electron + React + TypeScript) that lets you manage multiple Git worktrees and run a separate large‑language‑model (LLM) agent in each one. It’s built for developers who want to keep several feature branches alive at the same time while having a dedicated AI assistant (Claude, Gemini, Codex, etc.) attached to each branch.

Why it matters – Traditional IDEs treat a repository as a single workspace, so switching between branches means losing the context of any AI chat you had. EnsoAI makes each worktree a first‑class workspace with its own persistent AI session, terminal, and editor state, enabling true parallel development and AI‑assisted code review.


Core Features

Feature What you get
Multi‑Agent Matrix Choose from built‑in agents (Claude, Codex, Gemini, Cursor, Droid, Auggie) or add any CLI‑based agent. Each worktree keeps its own conversation history.
Git Worktree Management Create, switch, and delete Git worktrees from a visual sidebar; no manual git worktree commands needed.
Integrated Monaco Editor Multi‑tab code editing with syntax highlighting for 50+ languages, drag‑and‑drop reordering, and state persistence.
AI‑Powered Code Review Generate commit messages and run deep code reviews directly in the app using the selected agent.
3‑Way Merge Tool Visual conflict resolution with live preview of the merged result.
Visual Git UI Stage/unstage files, browse history, view diffs, all keyboard‑centric.
IDE Bridge One‑click “Open In” to jump to VS Code, Cursor, Ghostty, etc., while keeping the EnsoAI worktree context.
Multi‑Window & Theme Sync Open several workspaces at once; UI theme follows your terminal theme (400+ Ghostty themes).
Keyboard‑First Palette (Cmd+Shift+P), window switching (Cmd+1‑9), terminal shortcuts, etc.

Installation

  • macOSbrew tap j3n5en/ensoai && brew install --cask ensoai
  • Windows – via Scoop (scoop bucket add ensoai … && scoop install ensoai) or Winget (winget install J3n5en.EnsoAI)
  • Linux – AppImage, .deb, or build from source.
  • From source – clone the repo, pnpm install, then pnpm dev (dev mode) or pnpm build:<platform> for a packaged binary.

Tech Stack

  • Electron (desktop shell)
  • React 19 + Tailwind CSS 4 (UI)
  • TypeScript 5.9 (type‑safe code)
  • Monaco Editor (VS‑Code‑like editing)
  • xterm.js + node‑pty (embedded terminal)
  • simple‑git (Git operations)
  • sqlite3 (local settings / session storage)

Typical Workflow

  1. Open a repository in EnsoAI.
  2. Create a new worktree for a feature/bug.
  3. Launch the desired AI agent; the chat lives in that worktree’s folder.
  4. Edit files with the built‑in editor or open the folder in your favourite IDE.
  5. Use the terminal or the AI to run tests, generate code, or write commit messages.
  6. Commit, merge, or delete the worktree when the task is done.

Who should use it?

  • Parallel developers who need several feature branches active simultaneously.
  • Teams that rely on AI assistants for code generation, review, or documentation.
  • Experimenters who want a sandbox worktree for AI‑driven prototyping without polluting the main codebase.

License

MIT – free for personal, academic, or commercial use.


Bottom line – EnsoAI is a lightweight, AI‑centric workbench that augments Git worktree workflows with persistent, per‑branch LLM sessions, making it easier to develop, review, and experiment across multiple branches in parallel.

Related

  • Project
  • Project
  • Project
  • Project
  • Project