How is Agent Hub different from the Claude Code terminal CLI?

Same claude binary underneath — Agent Hub spawns it and passes stdin/stdout through. The difference is the UI: multiple threads, side-by-side approval, image paste, diff preview, SSH target switching, and slash-command fuzzy search.

Claude Codenative macOS GUI · open source · MIT

The Claude Code GUI you'll actually enjoy using.

Agent Hub is an open-source native macOS app that wraps the Claude Code CLI in a calm, keyboard-first chat window. Multiple threads, side-by-side tool approval, image paste, diff previews, and first-class SSH — so you can run Claude Code on your laptop or any VPS you own without ever leaving the app.

Download for macOS Clone on GitHub

Context

What is Claude Code?

Claude Code is Anthropic's agentic coding CLI — the one that plans, reads files, writes edits, runs commands, and asks for approval along the way. It's excellent on the terminal. It's even better with somewhere nice to put all the threads.

Why a GUI

Terminal is great. A GUI is better for long sessions.

Claude Code is at its best on multi-step tasks that span hours. The terminal handles one thread at a time, asks for approvals inline, and loses screenshots to the void. Agent Hub was built for the way you actually work with Claude Code.

Parallel threads

One tab per task. Keep the refactor going while you let another thread draft tests. No split panes, no multiplexer gymnastics.

Side-by-side tool approval

See the exact Read, Write, Edit or Bash call with its arguments before you approve. Diff preview for edits. Keyboard shortcuts for allow/deny.

Paste images, drop files

Screenshots from the clipboard and files from Finder go straight into the thread. Useful the moment you have a Figma frame or a stack trace to share.

SSH target switcher

Point a thread at omar@prod-1.fly.dev and Claude Code runs there — not on your laptop. Same UI, remote execution.

Slash commands, fuzzy-found

Every Claude Code slash command (/plan, /review, /compact, /memory) is one fuzzy-search away. Keyboard-first, zero mouse.

Hot-swap Claude models

Start in opus-4.7, drop to sonnet-4.6 for cheap iteration, bounce back. Per-thread choice, live.

Comparison

Claude Code CLI vs Claude Code in Agent Hub.

Same agent binary underneath. Different daily experience.

Capability	Claude Code CLI	Agent Hub GUI
Same claude binary under the hood	✓	✓
Parallel threads (tabs)	—	✓
Visual tool-approval dialog with diffs	Inline text	✓
Paste images from clipboard	—	✓
Drop files from Finder	—	✓
Slash-command fuzzy finder	—	✓
Per-thread model switch	Restart CLI	✓ (live)
Run on a remote VPS over SSH	Manual ssh session	✓ (one click)
Per-agent permission profiles	Global config	✓ (per thread)
Terminal-in-a-tab alongside threads	Separate window	✓
MIT license, free, no telemetry	✓	✓

How it works

Agent Hub spawns the real Claude Code CLI.

This isn't a reimplementation. Agent Hub launches the official claude binary (local or over SSH), captures its stdout, and renders each message, tool call, and approval request in a native macOS UI. That means every Claude Code feature — CLAUDE.md, MCP servers, project scopes, memory files, slash commands — works exactly the same. If it works in the CLI, it works in Agent Hub.

Get started

Run Claude Code in Agent Hub in three minutes.

You need Agent Hub, the claude CLI, and an Anthropic account (or API key). If you already have the CLI installed, Agent Hub picks it up on launch.

# 1. install Claude Code CLI (if you don't already have it)$ npm install -g @anthropic-ai/claude-code$ claude login # 2. grab Agent Hub$ git clone https://github.com/OmarDadabhoy/agenthub.git$ cd agent-hub && npm install && npm start # 3. new thread → pick "Claude Code" → go

Questions

Claude Code GUI FAQ.

Do I need an Anthropic API key to use Claude Code in Agent Hub?

If you install the claude CLI and log in with your Claude account, Agent Hub inherits that session — no extra key needed. If you prefer to use an API key directly, drop it into the provider config and Agent Hub uses it only for Claude Code threads.

Does Agent Hub store my Anthropic API key?

Keys live in the OS keychain on the machine where the CLI runs — locally on your Mac, or on the remote host when you're using SSH mode. Agent Hub itself never transmits or logs keys, and there's no telemetry.

Which Claude models work with Agent Hub?

Every model the claude CLI supports. Today that includes claude-opus-4.7, claude-sonnet-4.6, and claude-haiku-4.5. Switch models per thread with the /model slash command or the model picker in the sidebar.

Does Agent Hub respect my CLAUDE.md project files?

Yes. Agent Hub launches the claude CLI in your project directory, so CLAUDE.md (and nested CLAUDE.md files) load exactly the same way they do on the terminal.

Can I run Claude Code on a VPS from Agent Hub?

Yes — point Agent Hub at any host you can SSH into (for example omar@prod-1.fly.dev) and it runs the claude CLI there. The thread, permissions UI, slash commands, and file approvals all work identically. Your Mac is just the front-end.

How does Agent Hub handle Claude Code's tool permissions?

The approval dialog shows the exact tool call (Read, Write, Edit, Bash) with its arguments before it runs. Per-agent settings let you choose ask (prompt every time), acceptEdits (auto-approve safe edits in-repo), or bypassPermissions (never ask).

How is this different from the Claude Code terminal CLI?

Nothing is different under the hood — Agent Hub spawns the same claude binary and passes stdin/stdout through. The win is the UI: multiple threads, side-by-side approval, image paste, diff preview, SSH target switching, and slash-command fuzzy search. Same agent, nicer workspace.

Ready when you are

Get the Claude Code GUI for macOS.

Free, open-source, MIT. Clone Agent Hub from GitHub, build from source, or grab the .dmg. Requires macOS 11+ on Apple Silicon · 183 MB.

Download for macOS Clone on GitHub

Also supports Codex, Hermes, OpenClaw, and any OpenAI-compatible endpoint. See the full feature tour →