OpenClaw

• Multi-channel assistant access: OpenClaw lets users message one assistant from multiple chat apps through a single gateway process.
• Agent-native workflows: The platform supports tool use, session memory, multi-agent routing, and isolated sessions by sender, workspace, or agent.
• Media and device support: OpenClaw supports images, audio, documents, browser-based control, and paired mobile nodes for camera, voice, and canvas-based workflows.
• Self-hosted deployment: The product is designed to run on user-controlled infrastructure and supports macOS, Linux, and Windows environments.
• Control UI and onboarding: OpenClaw includes a browser-based control interface for chat, configuration, and session management, along with guided onboarding and daemon installation.

• Audience: Developers, technical users, and operators who want a personal or team assistant that can run on their own infrastructure and respond through existing communication channels.
• Use Case: OpenClaw is used for persistent AI assistance across messaging apps, agent-driven task execution, remote interaction from mobile devices, and workflow automation that stays tied to a user-controlled runtime rather than a hosted assistant service.

• Runtime Model: Single gateway process serving channels, sessions, routing, and control surfaces.
• Platform Support: macOS, Linux, and Windows.
• Deployment Requirement: Node 24 is recommended, with Node 22.19+ supported for compatibility.
• Licensing: Open source under the MIT license.

OpenClaw is fundamentally designed to be a self-hosted, autonomous AI personal assistant. Its core purpose is to sit as an always-on "control plane" on local hardware or VPS, listening on messaging channels (WhatsApp, Slack, Signal, iMessage, etc.) and executing complex system automation, data retrieval, and local task execution via natural language instruction.

Beyond being a single assistant, OpenClaw contains an embedded agent runtime that supports multi-agent routing. Users can design, configure, and deploy distinct subagents with isolated session states, specific system rules, custom tools, and dedicated channels (e.g., routing development questions to a coding agent, and search tasks to a research agent).

And users can interact with OpenClaw almost exclusively through chat-based message streams. Its Gateway handles multi-channel integrations with messaging clients, translating chat prompts into actionable system executions.