AutoSwarm

AutoSwarm, from HKUDS, is an open-source orchestration layer that sits on top of existing agent harnesses and CLIs. Given a task, it selects suitable agents, shares context and memory between them, runs subtasks in parallel, and verifies and synthesizes the results into a single answer. The README lists supported agents including Codex, Claude Code, OpenCode, Cursor, GitHub Copilot CLI, Meta Muse Code, CodeBuddy, Hermes, Nanobot, DeepSeek Harness, Grok Build and Kimi Code.

It is installed as a lightweight plug-in or skill into an agent the user already uses, which then acts as the entry point and can help configure the other locally installed agents. The main host needs Skill support and local command execution; collaborating agents need compatible local CLIs. It can also be used through a command line, IDE extensions, and AutoSwarm Desk, a desktop app for Windows and macOS (Apple silicon and Intel) that bundles Python and provides a private assistant, shared conversations, and an approval review view.

The project has an accompanying paper on arXiv ("AutoSwarm: Automating Agent Swarms") and describes its coordination algorithms in an in-repo archive. It does not provide models itself, so results depend on the underlying agents and their own accounts, credentials and costs. The roadmap lists more integrations, better coordination and improved reliability, indicating the project is still early.