Agent Swarm
Agent Swarm is an open-source operating system for AI work that coordinates a team of AI agents. A lead agent breaks down incoming tasks from Slack, GitHub, email, and other channels, then delegates them to specialized worker agents running in isolated Docker containers. It solves the problem of fragmented AI workflows by providing shared memory, persistent identity, and orchestration so that delegated work compounds across sessions, making it suitable for companies aiming to become AI-native.
✨ Key features
- Lead/worker orchestration in isolated Docker containers
- Compounding memory and persistent agent identity
- Multi-channel task intake: Slack, GitHub, GitLab, email, Linear, Jira, API
- Workflow engine with human-in-the-loop approval gates
- Scheduled and recurring tasks via cron
- Harness and LLM agnostic: supports Claude Code, Codex, and more
🎯 Use cases
- Automate feature development from Slack requests to pull requests
- Run proactive customer support with scheduled reports
- Generate content and manage social media presence
- Monitor code health and trigger fixes from alerting tools
- Create self-documenting release reports and videos
⚠️ Good to know
The README notes that the repo evolves daily and is in active development, but no specific limitations are stated.
❓ FAQ
What is Agent Swarm?
Agent Swarm is an open-source operating system for AI work that coordinates a team of AI agents to handle tasks autonomously.
How does Agent Swarm handle tasks?
A lead agent receives tasks from various channels, breaks them down, and delegates to worker agents running in isolated Docker containers.
Which AI models or harnesses are supported?
Agent Swarm is harness and LLM agnostic, supporting Claude Code, OpenAI Codex, Devin, and any Agent Client Protocol agent.
Can Agent Swarm integrate with existing tools?
Yes, it integrates with Slack, GitHub, GitLab, Linear, Jira, email, and more for task intake and output.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.