OpenClaw

OpenClaw is an open-source AI assistant and agent platform designed for developers, power users, automation enthusiasts, and organizations that want an AI agent running under their own control. Instead of operating only inside a proprietary hosted chatbot, OpenClaw runs its Gateway on the user's computer, server, or VPS and connects that agent to communication channels such as WhatsApp, Telegram, Slack, Discord, Signal, Microsoft Teams, Google Chat, iMessage, and WebChat. Workspace data, sessions, and memory can remain on infrastructure controlled by the user.

The platform is model-agnostic and supports more than 35 model providers, including OpenAI, Anthropic, Google, and other hosted services. It can also connect to self-hosted inference systems such as Ollama, vLLM, SGLang, llama.cpp, and LM Studio. Different agents can use separate models, workspaces, sessions, tools, and routing rules.

OpenClaw goes beyond conversational AI through browser automation, command execution, web search, scheduled jobs, persistent memory, skills, plugins, MCP connections, and multi-agent routing. Agents can work with documents, images, audio, and video, transcribe voice notes, generate media through configured providers, and interact with external services. ClawHub provides an ecosystem for extending agents with additional skills and plugins.

The core platform is self-hosted rather than a conventional hosted SaaS subscription. Costs can instead come from connected AI model APIs, web-search services, speech providers, cloud hosting, or other external APIs. Local or self-hosted models can reduce external API usage. Because OpenClaw agents can receive significant tool permissions, official documentation recommends careful access controls, sandboxing, dedicated environments, and separation of personal and shared business agents.