DISCOVER THE FUTURE OF AI AGENTSarrow_forward

SwarmClaw

calendar_todayAdded Apr 24, 2026
categoryAgent & Tooling
codeOpen Source
TypeScriptNode.jsWorkflow AutomationElectron桌面应用Multi-Agent SystemModel Context ProtocolAI AgentsAgent & ToolingModel & Inference FrameworkAutomation, Workflow & RPAProtocol, API & Integration

Self-hosted multi-agent AI runtime with native MCP, structured orchestration, persistent memory, and 23+ LLM providers.

SwarmClaw is a self-hosted multi-agent AI runtime built with TypeScript/Node.js, featuring a Next.js frontend and Electron-based cross-platform desktop app. It natively integrates the Model Context Protocol (MCP) with stdio, SSE, and streamable HTTP transports, enabling per-agent injection and assignment of tools from any MCP server.

On the orchestration layer, SwarmClaw provides structured session capabilities including templated runs, branching, looping, parallel branches, explicit joins, and restart-safe state persistence. Built-in heartbeat loops, scheduling, and background tasks enable autonomous operation. Delegation supports offloading tasks to external Code Agent CLIs (Claude Code, Codex, Gemini, etc.) as well as native sub-agents.

The memory system employs a hybrid recall strategy (vector + graph traversal + logs + persistent documents) with automatic reflective memory, communication preference learning, personal profile and boundary memory, and significant event recording. The skill system supports fixed skills, SKILL.md imports, and automatic extraction of reusable skill drafts from real conversations for review and enrollment.

SwarmClaw includes 23+ built-in LLM provider adapters covering mainstream Code Agent CLIs (Claude Code, Codex, Gemini, Copilot, Cursor, Qwen Code, etc.) and cloud APIs (Anthropic, OpenAI, DeepSeek, Groq, Mistral, xAI, etc.), with custom endpoint compatibility. Connectors span Discord, Slack, Telegram, WhatsApp, Teams, Matrix, and more. Crypto wallet integration (Base/Solana/Ethereum) supports balance queries, transaction simulation, execution, and approval thresholds.

Typical scenarios include personal assistants, virtual companies (CEO→CTO/CFO/CMO/COO hierarchy), multi-role dev team collaboration, research bureau parallel search, content studio pipelines, multi-platform help desks, and distributed OpenClaw fleet deployments.

Installation: Desktop app (swarmclaw.ai/downloads), global install (npm i -g @swarmclawai/swarmclaw), Docker deployment, or hosted deployment (Render/Fly/Railway). Requires Node.js ≥ 22.6.0, default port localhost:3456.

Unconfirmed: SwarmDock P2P market protocol details, OpenClaw independent project details, ClawHub skill platform details, Hermes Agent standalone repository link.

Related Projects

View All arrow_forward

STAY UPDATED

Get the latest AI tools and trends delivered straight to your inbox. No spam, just intelligence.

rocket_launch