Deep Technical Analysis · For Everyone
This document takes you deep inside Clawdbot's internal workings. Whether you're a regular user wanting to understand the basics, or a developer looking to dive into the code, you'll find what you need here.
Understanding Clawdbot's core features in plain language
Imagine having a super-smart assistant that you can chat with through WhatsApp, Telegram, Discord, and other messaging apps you already use. Just send messages like you would to a friend, and it helps you with all kinds of tasks.
Unlike regular chatbots, Clawdbot can execute real actions on your computer: run commands, process files, send emails, control smart home devices, and more. It doesn't just "talk"—it "does."
Your data never leaves your device. Clawdbot runs locally on your computer, keeping your private data away from third-party servers. You have complete control over your data.
Whether you're using WhatsApp, Telegram, or Discord, you can talk to the same AI assistant. Messages sync automatically, so you can get help anytime, anywhere.
Use your AI assistant on the platforms you already know
Understanding how Clawdbot's core components work together
Sends message
WhatsApp/Telegram/...
Message Routing Hub
Intelligence Engine
Command/File Operations
Multi-channel Broadcast
Receives reply
The central hub for all messages. No matter which platform messages come from, they arrive here first, then get processed and distributed. Supports real-time bidirectional communication.
The brain. Uses Anthropic's Claude model to understand your intent, decide what actions to take, and generate intelligent responses.
The "hands" that execute actual operations. Can run shell commands, read/write files, call APIs, control browsers, and more. Runs in a secure sandbox.
Converts unified message format to platform-specific formats. Each chat platform has a dedicated adapter handling connections and message exchange.
Understanding Clawdbot's technical implementation
Command-line interface entry point with 100+ subcommands for configuring and managing Clawdbot.
Core communication hub handling all client connections and message routing.
AI agent execution engine integrating Anthropic Claude API and tool systems.
Multi-platform message routing supporting 9+ built-in channels and 20+ extension plugins.
JSON5 format configuration management with Zod type validation.
Control panel frontend built with Lit Web Components.
How 390,000 lines of code across 3,015 files are organized
| Key File | Description | Responsibility |
|---|---|---|
| /src/entry.ts | CLI Entry Point | Node.js execution entry, environment initialization |
| /src/gateway/server.impl.ts | Gateway Server | WebSocket connection management, RPC handling |
| /src/agents/pi-embedded-runner.ts | Agent Runner | AI agent execution engine, tool scheduling |
| /src/web/auto-reply.impl.ts | Auto Reply | Message monitoring, reply generation and delivery |
| /src/channels/dock.ts | Channel Registry | Channel lifecycle management, start/stop |
| /ui/src/ui/app.ts | Frontend Main Component | Lit app entry, state management |
| /src/gateway/server-methods/chat.ts | Chat RPC | Chat message processing, history retrieval |
The complete journey from sending a message to receiving a reply
User sends a message through WhatsApp, Telegram, or other platforms. Messages can be text, images, voice, and other formats.
Channel adapter converts platform-specific message format to unified format, sends to gateway server via WebSocket.
Gateway's chat.send method validates message, processes attachments, loads session state, then starts Agent execution.
Pi Agent runtime calls Claude API, understands user intent, decides whether to invoke tools, generates response content.
If AI decides to perform operations, tool system runs commands, reads/writes files, or calls APIs in a secure sandbox.
AI response streams in real-time, gateway broadcasts delta events, frontend progressively displays reply content.
Final reply sent back to original platform via channel adapter, user receives reply in WhatsApp/Telegram app.
Technologies and tools used to build Clawdbot