Dielldev/egant

One native window for all your coding agents.
Claude Code, Codex, opencode and more — running in your real project folders,
with the files, diffs, terminals and git you need right beside the chat.
Download
**macOS (Apple Silicon):** download egant, open the DMG and drag egant to Applications. egant isn't notarized by Apple yet, so the first time you open it macOS says it can't verify the app: go to **System Settings → Privacy & Security** and choose **Open Anyway**. After that it keeps itself up to date (see Updates). egant drives the agent CLIs you already have, so install at least one first (for example claude, then claude login).
Intel Macs and Linux: build from source, see the Quickstart.
What is egant?
Coding agents live in terminals. That works, until you have five of them going across three projects and you want to see what they changed, review a diff, run the app, and pick up from your phone.
egant is a desktop app (Tauri: Rust + React) that wraps the agent CLIs you already have installed. You keep your own logins, your own git, your own gh. egant adds the workspace around them.
What you get
Chat with any agent
Start a session from the launch screen, pick an agent and model, and talk. Replies stream in token by token, tool calls show up as cards, and a subagent's work nests under the task that launched it. Interrupt mid-turn, switch permission modes, and answer permission prompts and questions inline.
A workspace around the chat
Press ⌘J for a panel with a file tree, git changes, pull requests and real terminals. Open files and diffs as tabs next to the conversation, with a syntax-highlighted editor and autosave.
Isolated by default
Each session can start in its own **git worktree** on its own branch, so parallel agents never step on each other or on your checkout.
Review and undo
See exactly what changed in a working-tree diff (unified or split), commit and push from the app, and **revert a turn** with per-turn snapshots when an agent goes the wrong way.
Find anything
⌘K searches across conversations and jumps to the exact message. @ mentions a file in the composer and / opens the command menu.
One library for every agent
Add MCP servers and skills once in the **Library**. egant syncs them into each agent's own config, so you stop maintaining them per tool.
A conversation next to the editor, with the file tree in the panel.
Review the working-tree diff, or preview the site the agent is building without leaving the window.
Agents
egant finds the agent CLIs on your machine and shows which ones are logged in. It does not ship or proxy any model.
| Kind | Agents | How it runs |
|------|--------|-------------|
| **Chat** | Claude Code, Codex, opencode | A native transcript, composer, permissions and usage meter, driven over each agent's own protocol |
| **CLI** | Cursor, Devin, Grok, Hermes, Pi, Copilot, Goose, Qwen and others | The agent's full terminal UI in a real PTY on the stage |
Chat sessions get the full egant experience. CLI sessions give you the agent as-is, with egant's worktrees, panel and project handling around it.
Your phone, too
The desktop app serves a small web client you can open on your phone, over your Tailscale network. Pair it from Settings → Devices with a QR code. Start a chat, switch models, and follow a running session while you are away from your desk. The agents keep running on your computer; the phone is just another window onto it.
Connect your phone to your computer
Do this on the computer that runs egant:
1. **Install Tailscale and sign in.** It's the only thing that carries your phone's connection to egant; egant itself only ever listens on 127.0.0.1.