crystal
Basic Information
Crystal is an Electron desktop application for developers to run, inspect, and test multiple Claude Code AI sessions concurrently using isolated git worktrees. It creates sessions from prompts with each session mapped to a separate branch and worktree, records each AI iteration as a commit, and lets you resume conversations and review changes. The app initializes a git repository if needed and requires Claude Code (installed or via API key) and git. It is an independent open source project by Stravu, designed to integrate with third-party deployments and cloud providers via environment settings, and is not affiliated with Anthropic. The project includes build and packaging scripts for development and macOS distribution.
Links
Stars
1280
Github Repository
Categorization
App Details
Features
Crystal provides parallel sessions so multiple Claude Code instances can run at once and each session is isolated in its own git worktree. It persists sessions so conversations can be resumed. Built-in git integration includes rebase from main, squash-and-rebase to main, and previewed commands. A diff viewer and change tracking surface modifications made by AI iterations. Desktop notifications alert when sessions need input. Project run scripts can be configured to test changes without leaving the app. The README documents prerequisites, build-from-source commands, production build scripts, and guidance for using cloud providers or corporate deployments via environment configuration.
Use Cases
Crystal helps developers iterate on AI-assisted code by keeping each experimental session isolated, versioned, and reversible so you can compare approaches and revert as needed. Automatic commits for each iteration make it easy to review diffs, perform manual edits, and then squash changes into a clean commit history before rebasing to main. Integrated run scripts enable testing changes immediately inside the project context. Session persistence and notifications reduce friction when managing many concurrent agent sessions. Support for third-party deployments and a configurable data directory make it practical for teams and for safe local development of the tool itself.
-
$$$$ -
$$$$