#OpenClaw Introduction

#What is OpenClaw?

OpenClaw is an open-source, self-hosted autonomous AI agent (MIT license) that became wildly popular after its January 2026 release. It runs on your own machine (Mac / Windows / Linux, even a $5/month VPS) and takes commands through the chat apps you already use — WhatsApp, Telegram, Slack, Discord, iMessage, Signal — to perform real tasks.

Unlike assistants that depend on a cloud SaaS, OpenClaw keeps its gateway, tools, and memory entirely local:

• Self-hosted / local-first: core operation doesn't rely on cloud services; your data stays with you
• Chat apps as the interface: command it from WhatsApp / Telegram like texting a friend
• Real execution power: shell commands, browser automation, email, calendar, file operations
• Text as memory: conversations, long-term memory, and skills are readable Markdown / YAML — Git-backable and greppable
• Can run autonomously: a built-in Heartbeat scheduler lets it work on its own without prompting

#History: Clawdbot → Moltbot → OpenClaw

The project was renamed several times; you may see different names online referring to the same lineage:

Clawdbot  →  Moltbot  →  OpenClaw (current official name)
(early prototype) (transitional)  github.com/openclaw/openclaw

Knowing this helps you avoid confusion when searching older materials.

#Design Philosophy: A "Control Plane" on Your Hardware

OpenClaw positions itself as a control plane running on your hardware: a persistent gateway process that unifies routing, tools, memory, and state.

Traditional cloud assistant       OpenClaw
───────────────────────────       ────────
Logic/data in the vendor cloud →  All inside your local process
Dedicated app / web entry      →  Reuses existing chat apps
Reactive only                  →  Proactive via Heartbeat
Black-box memory               →  Plain text files, readable & backable

#Core Features

#Who It's For

Great fit:

• You want a fully self-controlled, locally running private AI assistant
• You're used to firing off commands via WhatsApp / Telegram
• You need the AI to act proactively: scheduled checks, reports, reminders, automation
• You like plain-text, Git-able memory and skills (transparent, auditable)
• You're on a budget and want to run it long-term on a $5/month VPS

Be careful when:

• Powerful execution (shell / browser) carries real risk — security governance is mandatory (see Security)
• Community skills are not sandboxed by default; vet third-party skills carefully
• It iterates very fast and has had a critical vulnerability (patched) — keep it updated

#Relationship to & Comparison with Hermes Agent

OpenClaw shares lineage with Nous Research's Hermes Agent: Hermes is an MIT alternative built on OpenClaw's ideas, and the two can migrate between each other (hermes claw migrate).

Choosing:

• Prefer plain-text, lightweight, Node ecosystem, chat-driven → OpenClaw
• Want vector memory + self-improvement, Python ecosystem → Hermes Agent
• They're migratable — start on OpenClaw and switch smoothly later if you need stronger memory

#Key Facts

• License: MIT　First release: January 2026　Repo: openclaw/openclaw
• Requirements: Node.js 22+, minimum 2GB RAM / 2 cores
• Messaging platforms: WhatsApp, Telegram, Slack, Discord, iMessage, Signal, Mattermost (extension)
• Control panel: local Dashboard at http://127.0.0.1:18789/

#Next Steps

• Architecture & Features — how the single-process gateway works
• Installation & Usage — install and run your first chat
• Permissions & Security — read before you start