Summary

Open Swarm is a locally-running orchestrator designed to manage and coordinate multiple AI agents in parallel, addressing the complexity of juggling individual agent sessions. It targets developers and power users who need a unified interface to launch, monitor, and control swarms of agents, particularly for coding tasks. The primary benefit is enhanced productivity and centralized oversight through a single, intuitive dashboard.

How It Works

The system employs an Electron desktop shell housing a React/TypeScript frontend and a FastAPI backend. Core functionality revolves around a spatial, infinite canvas dashboard where agents are represented as draggable cards. Communication and control are managed via WebSockets, enabling real-time chat, unified tool-use approval workflows, and persistent conversation history. The architecture prioritizes a 100% local execution model, ensuring data privacy and eliminating reliance on third-party cloud backends.

Quick Start & Requirements

• Primary Install: Download the macOS Desktop App from GitHub Releases. Development setup requires cloning the repository (git clone https://github.com/openswarm-ai/openswarm.git), navigating into the directory (cd openswarm), and running bash run.sh.
• Prerequisites: Python 3.11+, Node.js 18+, Git. An Anthropic API key must be configured in the application's Settings page.
• Links: GitHub Releases

Highlighted Details

• Spatial Dashboard: An infinite canvas supports drag-and-drop agent cards, view cards, and embedded browser cards, allowing for customizable workspace layouts.
• Unified Approval Workflow: All tool-use requests from agents are surfaced in a single location for approval or denial, with options for batch actions.
• Message Branching: Users can edit prior messages to fork conversations and navigate between different conversational paths without losing context.
• Git Worktree Isolation: Each agent operates within its own isolated git worktree and branch, preventing conflicts between parallel tasks.
• Cost Tracking: Real-time tracking of estimated USD spend per agent session is provided.
• Interactive Views: Supports rendering interactive HTML/JS/CSS artifacts, backend Python execution, and agent-driven data visualization.

Maintenance & Community

No specific details regarding maintainers, community channels (e.g., Discord, Slack), or roadmap are provided in the README.

Licensing & Compatibility

• License: MIT License.
• Compatibility: The MIT license generally permits commercial use and integration into closed-source projects.

Limitations & Caveats

The desktop application is currently only available for macOS; Windows and Linux builds are planned but not yet released. Core agent functionality relies on an Anthropic API key, indicating a dependency on Anthropic's models.