Mozzie: Run Multiple Claude Code Agents in Parallel on Your Desktop

What It Does

Mozzie is a local-first desktop application that transforms Claude Code from a single-agent tool into a managed workforce. Instead of manually switching between tasks or running one Claude Code session at a time, Mozzie lets you describe what needs building in natural language, then automatically breaks it into work items, spins up isolated git worktrees, assigns Claude Code agents to each task, manages dependencies, and queues everything for your review.

The key innovation is parallel execution: Mozzie can run as many Claude Code agents simultaneously as your machine can handle, with each agent working on a separate task in its own isolated environment. All results funnel back to a single review interface where you can approve, modify, or merge changes.

Setup

Getting Mozzie running with Claude Code takes about 5 minutes:

git clone https://github.com/usemozzie/mozzie.git
cd mozzie
pnpm install
pnpm dev

Prerequisites: Node ≥20, pnpm ≥9, Rust (stable), and Claude Code installed locally.

Once the app opens:

1. Open Settings (gear icon) and add your Anthropic API key
2. Add Claude Code as an agent configuration
3. Open the command bar (Ctrl+K) and describe what you want built

When To Use It

Mozzie shines for three specific Claude Code workflows:

Parallel Feature Development: When you need to build multiple related features simultaneously. Instead of manually creating branches and switching contexts, describe the overall goal to Mozzie. It will create separate work items for each feature component, run Claude Code agents on each, and present unified diffs.

Large Refactoring Projects: For codebase-wide changes (like updating API patterns or migrating libraries), Mozzie can split the work across multiple agents. Each agent handles a different file or module, with Mozzie managing dependencies between them.

Multi-Repository Updates: When you need to make similar changes across multiple repositories (like updating configuration files or adding new dependencies), Mozzie can coordinate Claude Code agents across different codebases simultaneously.

How It Works With Claude Code

Mozzie treats Claude Code as one of several available "agents" (others include Gemini CLI and Codex). When you assign Claude Code to a work item:

1. Mozzie creates an isolated git worktree for that task
2. Spawns a Claude Code instance with the task description and context
3. Monitors progress and captures output
4. Presents the final changes in Mozzie's review interface

Because everything runs locally, there's no cloud dependency or data leaving your machine. This makes Mozzie particularly appealing for proprietary codebases where security is a concern.

Getting Started Tips

Start with small, well-defined tasks to understand Mozzie's workflow. The orchestrator works best when you provide clear, atomic requirements rather than vague descriptions.

Use the command bar (Ctrl+K) for quick task creation instead of manually configuring each work item. The LLM orchestrator (which can use OpenAI, Anthropic, or Gemini models) does the task breakdown for you.

Monitor system resources when running multiple Claude Code agents simultaneously. Each agent consumes memory and CPU; start with 2-3 parallel agents and scale up based on your machine's capacity.