swarmclaw: Run Self-Hosted AI Agent Swarms

Overview

SwarmClaw is a self-hosted AI agent runtime and multi-agent framework built with TypeScript for Node.js. It provides a control plane for running autonomous agents and agent teams, addressing the operational work of coordinating them: persistent memory, delegation, scheduled and background execution, tools, and operator controls.

It supports a broad range of model providers and agent runtimes, including Claude, GPT, Gemini, Ollama, and OpenClaw. The project is aimed at people who want to host and manage agent workflows themselves, from a personal assistant to a team of specialized agents. It can be installed as a desktop app or run as a server or Docker deployment.

Key Features

• Configure agents with different providers, models, personalities, and tool access.
• Coordinate agents through delegation, subagents, task execution, and structured sessions.
• Use durable memory, including hybrid recall, project context, and reflection features.
• Automate work with schedules, heartbeats, background jobs, and recovery controls.
• Connect MCP servers and messaging services such as Discord, Slack, Telegram, and WhatsApp.
• Manage OpenClaw gateway profiles and connect agents to external runtimes.
• Install as a desktop app on macOS, Windows, or Linux, or run the server through npm or Docker.

Use Cases

• A self-hosting individual can run a personal assistant that retains preferences and handles recurring tasks.
• A small engineering team can delegate implementation, testing, and review tasks to specialized agents.
• Researchers can run parallel agents to investigate separate sources or topics and synthesize findings.
• Operators managing several OpenClaw gateways can use one control plane to configure agents and monitor runtime status.

Getting Started

With Node.js 22.6 or later, install and start the server:

npm i -g @swarmclawai/swarmclaw
swarmclaw

Then open http://localhost:3456. See the README for desktop, Docker, configuration, and deployment instructions.

Alternatives

• nanobot: Nanobot is a Python personal-agent framework centered on chat interfaces, while SwarmClaw is a TypeScript runtime for persistent, scheduled multi-agent coordination.
• deer-flow: DeerFlow focuses on long-running research, coding, and content tasks with sandboxes, while SwarmClaw emphasizes persistent agents, schedules, and delegation.
• Memoh: Memoh gives each agent an always-on cloud computer with a desktop and browser, while SwarmClaw focuses on a TypeScript agent runtime and orchestration.
• deepagents: Deep Agents is a Python harness with built-in planning and file access, while SwarmClaw is a TypeScript runtime with persistent memory, schedules, and multi-agent coordination.

Considerations

The health snapshot describes the project as quiet: it reports no commits in the last 90 days, despite a recent release history that includes more than 100 releases in the past year. The latest listed release is v1.9.40. That contrast suggests a substantial release cadence but does not establish how quickly current bug reports will be handled; the snapshot also shows no issues closed or pull requests merged in the last 90 days. Thirteen contributors are recorded, but contributions are highly concentrated in one person.

Running from source or as an npm-installed server requires Node.js 22.6 or later. Docker is supported, but the README recommends Docker Desktop for sandbox browser execution. The project is MIT-licensed. The README documents many integrations and deployment paths, but this breadth also means adopters should verify the specific providers, connectors, and runtime workflows they depend on before committing to it.