# OpenClaw

> Personal AI assistant you self-host and talk to on the chat apps you already use

- Category: Agents & Automation / Assistants & Chatbots
- Language: TypeScript
- Repository: https://github.com/openclaw/openclaw
- Website: https://openclaw.ai
- Canonical page: https://ai-tldr.dev/tools/openclaw/
- Updated: 2026-09-19

## In short

OpenClaw is an open-source personal AI assistant that you run on your own devices instead of a hosted service. It connects to the messaging apps you already use, so you can chat with your assistant from WhatsApp, Telegram, Slack, Discord, Signal, iMessage, and many other channels. It sits in the Assistants & Chatbots category of the Agents & Automation tools and is open-source, written in TypeScript.

## Overview

OpenClaw is an open-source personal AI assistant that you run on your own devices instead of a hosted service. It connects to the messaging apps you already use, so you can chat with your assistant from WhatsApp, Telegram, Slack, Discord, Signal, iMessage, and many other channels.

At the center is a local-first Gateway that acts as the control plane: it manages sessions, channels, tools, and events. On top of that, OpenClaw can speak and listen on macOS, iOS, and Android, and can draw a live Canvas you control. It is built as an always-on assistant that feels local and fast, and since OpenClaw 2.0 a small team can share it through shared cloud sessions and operator roles.

Because it links to real messaging surfaces, OpenClaw treats inbound DMs as untrusted input. By default unknown senders get a pairing code and must be approved before the assistant will reply, which keeps a self-hosted bot from answering just anyone.

## What it does

- Multi-channel inbox covering WhatsApp, Telegram, Slack, Discord, Google Chat, Signal, iMessage, Matrix, Microsoft Teams, and many more
- Local-first Gateway as a single control plane for sessions, channels, tools, and events
- Multi-agent routing that maps channels, accounts, or peers to isolated agents with their own workspaces and sessions
- Voice Wake and Talk Mode for wake words on macOS/iOS and continuous voice on Android
- Live Canvas, an agent-driven visual workspace you can control
- Pairing-based DM security that blocks unknown senders until you approve them
- Shared cloud sessions with team operator roles and a shared credential store (OpenClaw 2.0)

## Getting started

OpenClaw runs on Node 24 (recommended) or Node 22.19+, and works with npm, pnpm, or bun. The recommended path is to install the CLI globally and run the guided onboarding, which installs the Gateway daemon so it stays running.

1. Install the OpenClaw CLI

   Install OpenClaw globally with your package manager. This gives you the openclaw command used for setup and day-to-day control.

```bash
npm install -g openclaw@latest
# or: pnpm add -g openclaw@latest
```

2. Run onboarding and install the daemon

   Start the guided onboarding to set up the gateway, workspace, channels, and skills. The --install-daemon flag installs the Gateway as a user service so it keeps running in the background.

```bash
openclaw onboard --install-daemon
openclaw gateway status
```

3. Send a test message

   Once the gateway is running, send a message through a connected channel to confirm everything works.

```bash
openclaw message send --target +1234567890 --message "Hello from OpenClaw"
```

4. Talk to the assistant

   Ask the assistant a question directly from the CLI; the reply can be delivered back to any connected channel such as WhatsApp, Telegram, or Slack.

```bash
openclaw agent --message "Ship checklist" --thinking high
```

## When to use it

- Run a private, always-on personal assistant on your own devices and reach it from WhatsApp, Telegram, or Slack
- Use voice on your phone or Mac, with wake words and continuous Talk Mode, to ask questions hands-free
- Route different chat accounts or contacts to separate, isolated agents with their own workspaces
- Keep a self-hosted bot safe by requiring pairing approval before it responds to unknown senders

## Version history

- 2026-09-19 v2026.9.5: Plugins install without restarting the Gateway, and updates are atomic — the installer checks the next version before switching over. Adds GPT Live for meetings and phone calls, read-only conversation sharing, shared browser pages, conversation archiving, and a guided setup for specialist agent teams. 4,179 pull requests from 502 contributing accounts.
- 2026-09-11 v2026.9.4: Plugin and skill management moves into the Control UI: one workspace discovers bundled and ClawHub plugins, installs them, and shows which skills were learned from past conversations. A compatible failed update can now roll back to the previous packages and their configuration. The `OPENCLAW_CLAUDE_CLI_LOG_OUTPUT` alias is removed — use `OPENCLAW_CLI_BACKEND_LOG_OUTPUT` instead.
- 2026-09-05 v2026.9.2: Adds GPT-6 Astra (as `openai/gpt-6-astra`) and Muse Spark 1.3, with async function tools and mid-turn steering on the OpenAI Responses endpoint. Swarm sub-agents are on by default, and most settings apply without restarting the Gateway. Session tools now default to all-session visibility — set `tools.sessions.visibility` to `agent` or `self` on a shared Gateway before upgrading.
- 2026-08-30 v2026.8.1 (OpenClaw 2.0): The largest release in the project's history: over 16,000 pull requests from 933 contributors add conversation search, shared cloud sessions with team operator roles, self-learning skills, and a rebuilt browser experience. The bundled OpenProse plugin is removed and codex/* model references move to openai/*, so run `openclaw doctor --fix` after upgrading.
