OpenClaw Reference

OpenClaw is an open-source autonomous AI personal assistant created by Peter Steinberger in late 2025. This document serves as a reference for understanding its architecture and features, which can inform the design of lsbot.

Overview

OpenClaw is a self-hosted agent runtime and message router that acts as a personal AI assistant running on your own machine. It went viral with 100k+ GitHub stars and 2M visitors in one week.

Originally released as "Clawdbot", then renamed "Moltbot" (after Anthropic's trademark request), and finally "OpenClaw" in early 2026.

Architecture

┌─────────────────────────────────────────────────────────┐
│                    OpenClaw Runtime                      │
│                    (Node.js Service)                     │
├─────────────────────────────────────────────────────────┤
│                                                          │
│  ┌──────────────┐  ┌──────────────┐  ┌──────────────┐   │
│  │   Message    │  │   Agentic    │  │  Persistent  │   │
│  │   Router     │  │    Loop      │  │   Memory     │   │
│  └──────┬───────┘  └──────┬───────┘  └──────┬───────┘   │
│         │                 │                 │            │
└─────────┼─────────────────┼─────────────────┼────────────┘
          │                 │                 │
          ▼                 ▼                 ▼
┌─────────────────┐  ┌─────────────┐  ┌─────────────────┐
│ Chat Platforms  │  │  External   │  │  Local Machine  │
│ - WhatsApp      │  │  AI Models  │  │  - Files        │
│ - Telegram      │  │  - Claude   │  │  - Apps         │
│ - Discord       │  │  - GPT      │  │  - Calendar     │
│ - Slack         │  │  - etc      │  │  - etc          │
│ - iMessage      │  │             │  │                 │
│ - Signal        │  │             │  │                 │
│ - Teams         │  │             │  │                 │
└─────────────────┘  └─────────────┘  └─────────────────┘

How It Works

Self-hosted agent runtime - Runs as a long-running Node.js service on your machine
Multi-channel messaging - Connects to WhatsApp, Telegram, Discord, Slack, iMessage, Signal, Teams, Matrix, Google Chat, etc.
Agentic loop - Executes tasks autonomously, not just conversational chat

Key Features

Feature	Description
Persistent memory	Remembers context across sessions, learns patterns over time
Local machine access	Deep integration with your local apps and files
Heartbeat	Proactively wakes up and takes action without user prompting
Multi-platform inbox	One AI assistant across all messaging platforms
Contextual understanding	Learns relationships (e.g., recognizes work emails vs personal)

Core Capabilities

Three Pillars

OpenClaw works because it does three simple things:

Persistent memory across sessions - Retains context between conversations
Deep local machine access - Unapologetic access to your apps and files
Autonomous agentic loop - Takes action, not just suggests steps

Supported Platforms

Messaging:

WhatsApp
Telegram
Discord
Slack
iMessage
Signal
Microsoft Teams
Matrix
Google Chat

Productivity Apps:

Apple Notes
Apple Reminders
Things 3
Notion
Obsidian
Trello
GitHub

Use Cases

Developer & Technical Workflows

Automate debugging and DevOps tasks
Codebase management with GitHub integration
Scheduled cron jobs and webhook triggers

Personal Productivity

Manage calendars and reminders
Cross-app task management from a single conversation
Research and information gathering

Autonomous Actions

Notable example: When unable to make a restaurant reservation through OpenTable, OpenClaw obtained AI voice software and called the restaurant directly to secure the reservation.

Security Considerations

OpenClaw explicitly trades security for capability. From their FAQ:

"There is no 'perfectly secure' setup."

Recommendations:

Best used on personal devices
Not recommended for enterprise environments without proper guardrails
Consider using a separate device for OpenClaw

Relevance to lsbot

Features to consider adopting:

Multi-channel support - Currently lsbot uses MCP stdio, could expand to messaging platforms
Persistent memory - Add session memory across tool calls
Heartbeat/proactive actions - Scheduled tasks without user prompting
Contextual learning - Learn user patterns over time