Deploy Clawdbot: Turn Your Computer into a Local‑First AI Personal Assistant
This article introduces Clawdbot, a local‑first AI agent that runs on your machine, explains its architecture and unique product positioning, provides step‑by‑step deployment instructions (including environment setup, cloning, configuration, and launching), and discusses advanced skill integration, security risks, and broader implications for AI‑driven workflows.
Project Overview
Clawdbot (renamed Moltb) is an open‑source personal AI assistant that runs locally on macOS, Linux, or WSL2. It follows a local‑first architecture, interacts via existing chat platforms (Telegram, Slack, iMessage, Teams) and can manipulate local files, terminals, and browsers. Source code: https://github.com/clawdbot/clawdbot
Prerequisites
Node.js v20 or newer (project written in TypeScript).
Bun (optional, faster installation).
LLM API key for Anthropic (Claude), OpenAI, or Gemini.
Telegram bot token (recommended communication channel).
Quick Start
1. Clone and install
git clone https://github.com/clawdbot/clawdbot.git
cd clawdbot
# install dependencies
bun install # or npm install2. Configure environment
Create a .env file in the repository root with at least the following entries (example for Claude):
# Model configuration
ANTHROPIC_API_KEY=your_api_key_here
# Telegram configuration
TELEGRAM_BOT_TOKEN=your_bot_token
TELEGRAM_ALLOWED_USER_IDS=your_telegram_user_id3. Run the agent
bun run dev # starts the local agent in development modeCommunication Gateway
Telegram is the simplest gateway. Steps:
Create a bot with @BotFather and obtain the token.
Insert the token into TELEGRAM_BOT_TOKEN in .env.
Start the agent and send ping to the bot; a pong reply confirms connectivity.
Other gateways (iMessage, WhatsApp) require the clawd-gateway tool and platform‑specific setup.
Extending Capabilities (Skills)
Clawdbot’s functionality is provided by the MCP skill system. Skills are located in the skills directory. To enable a skill, add its name to the configuration (e.g., fetch, filesystem). Example command from a mobile client:
“List all desktop files containing the word ‘Confidential’.”
Common Pitfalls
Storage : The agent creates a hidden directory ~/.clawd (or similar) to store memory and logs. Deleting this directory resets the agent’s state.
Network : The agent frequently calls the LLM provider’s API; a stable internet connection is required.
Security : Always set TELEGRAM_ALLOWED_USER_IDS (or the equivalent for other channels). Running the agent on a public server without this restriction can expose your API quota and file system.
References
GitHub repository: https://github.com/clawdbot/clawdbot
Documentation: https://docs.clawd.bot
Skill marketplace: https://clawdhub.com
Tencent Technical Engineering
Official account of Tencent Technology. A platform for publishing and analyzing Tencent's technological innovations and cutting-edge developments.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.
