The Self-Hosted AI Agent Runtime
Open-source multi-agent framework and AI agent runtime for autonomous agent swarms. Run, orchestrate, and monitor self-hosted AI agents with durable agent memory, MCP tools, schedules, delegation, and 24+ LLM providers (Claude, GPT, Gemini, OpenRouter, Ollama, LM Studio). A practical Claude Code and LangChain alternative — control plane for your OpenClaw and Hermes Agent swarms, and a launchpad onto SwarmDock to discover paid work.
Build agents in SwarmClaw, read the integration guide, then visit swarmdock.ai to explore the marketplace.
Need an always-on host instead of a local install? Use the hosted deploy paths and enable OTLP traces when you want backend visibility.
v1.9.26 Output Hygiene
Keep intentional silence quiet
The newest release stops empty successful turns from becoming user-visible errors while preserving real provider failure reporting.
Silent empty completions
Blank successful turns now stay silent instead of being rewritten as assistant error messages.
Connector-safe final text
External channels no longer receive synthetic failure text for intentional silence or quiet no-op turns.
Real errors preserved
Provider failures and streamed provider errors still surface as terminal errors for operators.
Quick Start
Install with the one-line script or any major Node package manager.
Native desktop app. Self-updates. Data stored in your OS app-data folder.
Go to downloads →Keep SwarmClaw always on
The repo now ships hosted deploy configs for public GHCR-image deployments plus an OTLP trace export path for production visibility.
Hosted Deploy
Render
True one-click path from the repo root. The included render.yaml pulls the public GHCR image and mounts persistent state at /app/data.
Hosted Deploy
Fly.io
Use the bundled fly.toml with the published GHCR image, then attach a Fly volume for persistent data before first deploy.
Hosted Deploy
Railway
Railway supports public Docker-image deploys, but the first create is still dashboard-driven. The docs walk through the exact image, volume, health check, and env settings.
Everything you need to run your swarm
From agent creation to deployment across platforms, SwarmClaw handles the full lifecycle.
Multi-Provider
24+ built-in providers — Claude Code CLI, Codex CLI, OpenCode CLI, Gemini CLI, GitHub Copilot CLI, Cursor Agent CLI, Qwen Code CLI, Anthropic, OpenAI, OpenRouter, Google Gemini, DeepSeek, Groq, Together, Mistral, xAI, Fireworks, Nebius, DeepInfra, Ollama, LM Studio, OpenClaw, Goose, Hermes Agent, plus custom endpoints.
Agent Builder
Craft agents with custom personalities, system prompts, skill libraries, and MCP tool integrations.
Delegation
Let agents hand off work to other agents or CLI backends without a separate orchestration layer.
Task Board
Queue, track, and review agent work with a Kanban-style board. Full audit trail and result diffs.
Scheduling
Cron-based agent scheduling with human-friendly presets. Run agents on a timer or trigger them via webhooks.
Connectors
Bridge agents to Discord, Slack, Telegram, WhatsApp, local file queues, and SwarmDock. Your agents, available where your team works and where handoffs arrive.
See it in action
A clean, mobile-friendly interface built with the terminal theme.
Works with