Run and watch Claude Code agent sessions in a native macOS GUI instead of the terminal.
Browse and resume past Claude Code sessions with a live conversation view.
Give a Claude agent a local, GPU-backed image generation tool it can call during a session.
Split the window into multiple panes to run several agent sessions side by side.
| s-age/gazzo | aiduckman/claudeusage_latest_may2026 | arnabau/thermalpulse | |
|---|---|---|---|
| Stars | 0 | 0 | 0 |
| Language | Swift | Swift | Swift |
| Setup difficulty | hard | easy | moderate |
| Complexity | 4/5 | 2/5 | 3/5 |
| Audience | developer | vibe coder | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires macOS 15+, Swift 6, the Claude CLI, and a separate build step to enable the local image generation engine.
Gazzo is a macOS desktop app that gives Claude Code, an AI coding agent tool, a graphical interface instead of only running it in a terminal. It lets a user browse past Claude Code sessions stored on their machine, start new agent sessions, resume old ones, and watch the conversation unfold live, including the agent's tool use and any subagents it spawns. Users can split the window into multiple panes, each running its own session, pick which AI model and working directory to use, and mention files directly in the composer with search. If a headless Claude session needs to ask the user something, gazzo can show that as a native prompt instead of the session failing silently. The project doubles as an experiment in giving an AI agent the ability to generate images on its own. It runs a local image generation engine on the Mac's GPU and exposes it to Claude sessions as a callable tool, alongside tools that let the agent ask the user for permission or ask them to choose between options. The image engine runs inside a separate helper process so it keeps working even when the main window is closed, and generated images are saved to disk and shown in an artifact viewer with a live preview. The author notes that once given this tool, Claude uses it on its own, for example illustrating an explanation or asking which of several generated options the user prefers. Under the hood, gazzo is built on a custom Swift framework called swift-kernelee, using an architecture where every action across the app is represented as data flowing through a central kernel, and the interface only reads from a shared state buffer rather than writing to it directly. The app never modifies a user's original Claude Code history files, keeping its own separate data store instead. Building it requires macOS 15 or newer, Swift 6, and the Claude CLI installed, and image generation requires an extra build step to compile the local image engine. The project is released under the MIT license.
A macOS GUI for driving Claude Code AI agent sessions, with an experimental local image-generation tool the agent can call on its own.
Mainly Swift. The stack also includes Swift, SwiftUI, Metal.
Use freely for any purpose, including commercial use, as long as you keep the copyright notice.
Setup difficulty is rated hard, with roughly 1h+ to a first successful run.
Mainly developer.
This repo across BitVibe Labs
double-check against the repo, no cap.