openclaw: A personal AI assistant that runs on your own devices
Project Overview
In a landscape where AI assistants are increasingly tethered to cloud services and corporate data policies, OpenClaw represents a deliberate counter-movement. The project, which has accumulated over 369,000 stars on GitHub[1], positions itself as a personal AI assistant that runs on your own devices rather than someone else’s servers. What makes this project architecturally interesting is its “Gateway” model — a lightweight control plane that routes between your chosen LLM backend and an extensive array of messaging channels. The crustacean branding and “EXFOLIATE! EXFOLIATE!” tagline are playful, but the underlying design philosophy is serious: the Gateway is intentionally minimal, serving only as the orchestration layer while the actual intelligence comes from whichever model provider you plug in. This separation of concerns means you’re not locked into a specific AI provider, and your conversation history stays on infrastructure you control. The project’s sponsorship by OpenAI, GitHub, NVIDIA, and Vercel[2] signals institutional confidence, though the architecture itself remains provider-agnostic.
What It’s For
OpenClaw solves a specific pain point for developers and privacy-conscious users who want AI assistant capabilities across multiple communication platforms without centralizing their data in a third-party service. If you’ve ever wanted a single AI assistant that can respond in both your Discord server and your WhatsApp group chats, while maintaining consistent context and memory, this is the tool for that. The supported channels list reads like a directory of modern messaging — WhatsApp, Telegram, Slack, Discord, Signal, iMessage, IRC, Microsoft Teams, Matrix, and over a dozen more[3]. This breadth is both the project’s strongest differentiator and its most significant complexity burden. Where this project really shines is for users who already self-host services or are comfortable with Docker/Nix setups. The tradeoff is clear: you trade the convenience of a managed service for data sovereignty and the ability to use your choice of LLM backend. Teams or organizations that need shared, multi-user assistant access should look elsewhere — this is explicitly designed for single-user personal use.
How to Use It
The primary entry point is the openclaw onboard command, which runs an interactive setup wizard guiding you through configuring the gateway, workspace, channel connections, and skills. This is a thoughtful design choice — rather than presenting a wall of configuration files, the onboarding process walks you through each decision point sequentially. After onboarding, the assistant runs as a background process, listening on whatever channels you’ve configured. The real power emerges from the skill system, which allows you to extend the assistant’s capabilities beyond simple chat. Skills can range from web searching to custom API integrations, and they’re managed through the same CLI tooling. The project supports npm, pnpm, or bun as package managers, and recommends WSL2 for Windows users to avoid the compatibility issues that plague cross-platform Node.js services.
Interactive setup wizard that guides through gateway, workspace, channel, and skill configuration
openclaw onboardLaunches the assistant as a background process, connecting to configured channels
openclaw startInstalls a new skill to extend the assistant’s capabilities beyond base chat
openclaw skill add <name>Recent Updates
Latest Release: 2026.5.6 (2026-05-06)
Latest stable release in the 2026.5.x series, continuing the project’s calendar-versioning scheme
The project follows a rapid release cadence with multiple patches per month, suggesting active development and a responsive maintainer team. The calendar versioning (YYYY.MINOR.PATCH) provides clear temporal context for releases. The Discord community with over 145,000 members[4] indicates healthy community engagement around self-hosting and configuration support.
Sources & Attributions
[1] 369,217 stars on GitHub as of repository data — openclaw/openclaw [2] Listed as sponsors in the project README — openclaw/openclaw [3] Supported channels listed in README include WhatsApp, Telegram, Slack, Discord, Google Chat, Signal, iMessage, BlueBubbles, IRC, Microsoft Teams, Matrix, Feishu, LINE, Mattermost, Nextcloud Talk, Nostr, Synology Chat, Tlon, Twitch, Zalo, WeChat, QQ, WebChat — openclaw/openclaw [4] Discord community size referenced in README badge — openclaw/openclaw