About This Site

This website (openclaw-ai.online) is a resource and documentation site for OpenClaw AI, a self-hosted personal AI assistant. This site provides comprehensive documentation, tutorials, guides, and reference materials to help users understand, install, configure, and use OpenClaw.

📚 What This Site Provides

This documentation site is not the OpenClaw project website. It is an independent resource site created to help users learn about and use OpenClaw AI.

For the OpenClaw project, including source code, issues, discussions, and community contributions, please visit:

OpenClaw (formerly known as Moltbot and Clawdbot) is a self-hosted personal AI assistant that enables you to run a powerful AI assistant on your own hardware. Key features include:

• Complete Data Privacy: All data stays on your devices
• Multi-Platform Support: Works on macOS, Windows, Linux, iOS, and Android
• Multiple Channels: Connect via WhatsApp, Telegram, Slack, Discord, Google Chat, Signal, iMessage, Microsoft Teams, WebChat, BlueBubbles, Matrix, Zalo, Zalo Personal, and more
• Extensible Skills: Create custom skills or use community-built ones
• Browser Automation: Navigate websites and automate web tasks
• Voice Features: Voice wake, push-to-talk, and text-to-speech support
• Persistent Memory: Long-term memory system that learns your preferences
• Workflow Automation: Automate complex tasks and workflows

OpenClaw is an open-source project licensed under the MIT License, developed by the OpenClaw community. The project is becoming an independent foundation and remains open source. It is not affiliated with Anthropic or other AI service providers.

Highlights from the current stable release (see also Releases & Changelog):

• Models: Bundled Codex provider—codex/gpt-* vs openai/gpt-* routing is explicit so Codex-managed auth and threads stay on the Codex path.
• Memory: Optional Active Memory plugin (sub-agent before the main reply). See the project's Active Memory concept doc.
• Ops: openclaw exec-policy to sync tools.exec.* with your local exec approvals file; commands.list RPC for remote gateway clients.
• Talk & channels: Experimental macOS MLX Talk provider; Teams message actions and many Teams/WhatsApp/media fixes; Matrix MSC4357 live typing markers on drafts.
• Security: Broad browser/sandbox SSRF and navigation hardening, exec preflight and env tooling, plus dependency and realtime WebSocket frame handling improvements.
• Recent trains: v2026.4.9 Matrix DM doctor --fix, dreaming UI, node exec sanitization. v2026.4.8 npm sidecar startup fixes. v2026.4.5 legacy config alias cleanup—always run openclaw doctor --fix after upgrading.

Still catching up: v2026.4.2 moved x_search and Firecrawl web_fetch into plugin-owned config (see xAI and Firecrawl guides).

Full changelog: v2026.4.10 on GitHub. All releases: Releases · openclaw/openclaw.

2026.2.26: External Secrets Management (openclaw secrets), ACP thread-bound agents, openclaw agents bindings/bind/unbind, Codex WebSocket-first, channel-plugin onboarding hooks, Gemini OAuth risk gate, Android device/notifications. Fixes: Telegram/DM allowlist, delivery queue backoff, Google Chat/Nextcloud Talk lifecycle, typing pipeline, Teams file uploads, compaction/onboard scope.

2026.2.25: Android chat streaming and markdown, heartbeat directPolicy (allow|block), onboarding security notices (personal-by-default), branding to ai.openclaw, gateway/auth security (pairing for shared-token operator, WebSocket origin checks, Control UI trusted-proxy). Subagent delivery, Telegram webhook, Slack threads, cron routing, model fallback, typing-indicator fixes.

2026.2.24: Auto-reply/abort stop phrases (multilingual), Android four-step onboarding and five-tab app, Talk/Gateway provider-agnostic config, security.trust_model.multi_user_heuristic, heartbeat DM block default (overridden by directPolicy in 2026.2.25), sandbox container-namespace block. Routing, Discord voice/streaming, WhatsApp/Matrix/Telegram/Android/Slack/Zalo/macOS fixes. Details in Installation, CLI, and Security.

This documentation site was created to provide:

Some projects are inspired by OpenClaw but use a different stack (e.g. Claude Code instead of the full OpenClaw Gateway). They can be useful if you want a minimal, channel-focused setup. This site does not endorse or maintain them; links are for discovery only.

While we strive to keep this documentation accurate and up-to-date, the source of truth for OpenClaw is the project’s GitHub repository. If you find any discrepancies or have questions, please refer to the repository or open an issue there.

Ready to start using OpenClaw? Here are some helpful resources: