What Is Moltbot?

An open-source, local-first personal AI assistant that runs on your own devices, connects to all your chat apps, and actually does things for you.

The Personal AI Assistant You Control

Moltbot is a personal, single-user AI assistant designed to feel local, fast, and always available. Unlike cloud AI tools that run on remote servers, Moltbot runs on your own machine or server.

It connects directly to your chat accounts and acts as your assistant — not a shared SaaS product. The Gateway is just the control plane. The product is the assistant.

"Moltbot isn't just another chatbot — it's a personal operating system for AI."

Core Concepts

Understanding the fundamentals of Moltbot

🏠

Local-First Architecture

Moltbot runs on your own hardware (Mac, Linux, Windows via WSL2, Raspberry Pi, or VPS). Your data and credentials never leave your control.

🚪

Self-Hosted Gateway

The Gateway is a background service acting as the front door, managing connections to messaging platforms and orchestrating all interactions.

👤

Single-User Design

Designed for personal use, not as a shared chatbot service. One assistant that knows you and works exclusively for you.

Always-On Availability

Runs as a daemon service 24/7, maintaining long-term memory and context. No need to be at your computer to interact.

System Architecture

Moltbot's architecture is centered on a persistent gateway service with four primary components working together.

WhatsApp / Telegram / Slack / Discord / Signal / iMessage / Teams / Matrix
                              │
                              ▼
              ┌───────────────────────────────┐
              │           Gateway             │
              │       (control plane)         │
              │    ws://127.0.0.1:18789       │
              └──────────────┬────────────────┘
                             │
         ┌───────────────────┼───────────────────┐
         │                   │                   │
         ▼                   ▼                   ▼
   ┌──────────┐       ┌──────────┐       ┌──────────┐
   │ Pi Agent │       │  Browser │       │  Canvas  │
   │  (RPC)   │       │ Control  │       │  (A2UI)  │
   └──────────┘       └──────────┘       └──────────┘
         │
         ├─ CLI (moltbot …)
         ├─ WebChat UI
         ├─ macOS App
         └─ iOS / Android Nodes

Four Primary Components

🌐

1. Gateway

The central control plane running on your machine. Manages WebSocket connections, session management, tool orchestration, security, authentication, and exposes a web-based Control UI.

🧠

2. Pi Agent

The AI brain and reasoning engine. Supports Claude, GPT, Gemini, and local models via Ollama/LM Studio. Handles thinking, tools, memory, and routing with RPC communication.

🧩

3. Skills

Modular capabilities that extend the agent's reach. 50+ bundled skills including browser automation, file system access, calendar integration, smart home control, and more.

💾

4. Memory

Persistent storage layer retaining context, preferences, and long-term conversation history. Uses Markdown files (USER.md, MEMORY.md) that grow as you interact.

Multi-Channel Messaging

Moltbot connects to all your existing chat apps through a unified inbox. One conversation view is maintained across platforms, so the assistant remembers context regardless of channel.

💬

WhatsApp

via Baileys

✈️

Telegram

via grammY

💼

Slack

via Bolt

🎮

Discord

via discord.js

📱

iMessage

via imsg CLI

🔒

Signal

via signal-cli

👥

Teams

extension

🌐

Matrix

extension

Plus Google Chat, BlueBubbles, Zalo, WebChat, and more via extensions.

Moltbot vs. Cloud AI Assistants

FeatureMoltbotCloud AI (ChatGPT, Claude.ai)
Data LocationYour device / serverProvider's servers
AvailabilityAlways-on 24/7Requires opening app/website
MemoryPersistent across all channelsLimited or session-based
Proactive ActionsCan message you firstReactive only
System AccessFull (shell, browser, files)Sandboxed / limited
Chat Platforms10+ messaging appsSingle interface
CustomizationFully configurableLimited settings
Model ChoiceAny provider + local modelsProvider's model only

Key Features

What makes Moltbot powerful

🧠

Persistent Memory

Unlike stateless AI, Moltbot remembers everything. Conversations on WhatsApp continue seamlessly on Telegram. The USER.md file grows as you interact.

Proactive Communication

Features a Heartbeat Engine and cron job integration. Moltbot can wake itself up, check data, and message you when thresholds are met.

🎤

Voice Wake + Talk Mode

Always-on speech support for macOS, iOS, and Android with ElevenLabs integration. Talk, listen, and get natural voice responses.

🎨

Live Canvas

Agent-driven visual workspace with A2UI. See live visual output you control, perfect for code, diagrams, and interactive content.

🌐

Browser Automation

Dedicated Chrome/Chromium control with CDP. Snapshots, actions, uploads, and profile management for web automation tasks.

🔀

Multi-Agent Routing

Route inbound channels, accounts, and peers to isolated agents with workspaces and per-agent sessions for complex workflows.

Extensible Skills System

Skills extend Moltbot's capabilities, allowing it to interact with external services, automate workflows, and perform specialized tasks. Moltbot comes with 50+ bundled skills that load automatically.

🏠 Smart Home

  • Home Assistant
  • Philips Hue
  • Google Nest
  • Nanoleaf
  • Alexa / VoiceMonkey

🔍 Search & Research

  • Brave Search
  • Kagi Search
  • Exa AI
  • Literature Review
  • SEO Audit

⚙️ Productivity

  • GitHub
  • Notion
  • Calendar
  • Email
  • n8n Workflows

Install more skills from ClawdHub:

clawdhub search "what you need"
clawdhub install <skill-name>

Technology Stack

Moltbot is built with modern technologies for reliability and extensibility.

Runtime & Core

  • Node.js ≥ 22 — Runtime requirement
  • TypeScript — Type-safe codebase
  • WebSocket — Real-time Gateway communication
  • RPC — Agent communication protocol

AI Model Support

  • Anthropic Claude — Primary recommended
  • OpenAI GPT — Full support
  • Google Gemini — Full support
  • Ollama / LM Studio — Local models

Channel Libraries

  • Baileys — WhatsApp Web
  • grammY — Telegram Bot API
  • discord.js — Discord Bot
  • Bolt — Slack integration

Deployment Options

  • npm / pnpm — Global install
  • Docker — Containerized deployment
  • Nix — Declarative config
  • systemd / launchd — Service management

Workspace Structure

Moltbot stores all configuration, memory, and context in a local workspace directory. These files are fully editable and allow you to customize how Moltbot behaves.

~/.clawdbot/
├── AGENTS.md        # Available agents
├── BOOTSTRAP.md     # Initial system setup
├── HEARTBEAT.md     # System health
├── IDENTITY.md      # Moltbot's identity
├── SOUL.md          # Personality traits
├── TOOLS.md         # Available tools
├── USER.md          # What it knows about you
├── canvas/          # Working directory
├── memory/          # Persistent memory
└── agents/
    └── <agentId>/
        └── agent/
            └── auth-profiles.json

💡 Tip: Edit USER.md to add context about yourself, your projects, and preferences. The memory/ directory stores long-term memories that persist across sessions.

Security Model

Moltbot connects to real messaging surfaces, so security is built in by default.

🔒

Local Loopback by Default

Gateway is bound to 127.0.0.1 by default, limiting access to the host machine only.

DM Pairing Required

Unknown senders must be approved before they can interact with your assistant. Explicit opt-in for public messages.

📋

Local Allowlists

Configure exactly who can interact with your assistant through allowFrom settings per channel.

🩺

Built-in Diagnostics

Run moltbot doctor to surface risky or misconfigured DM policies and update service entrypoints.

In Summary

Moltbot is fundamentally an orchestration layer rather than a standalone model. It combines:

  • A gatekeeper service with a database of context and credentials
  • An LLM endpoint of your choice for reasoning
  • Integrations (skills) for email, calendars, browsers, smart home, and more
  • Multi-channel bridges to all your messaging apps

The result is a personal, agentic AI system that combines powerful language models with tool integration and local execution — a capable, privacy-focused alternative to cloud-dependent assistants.