How to Install and Customize Your Personal AI: A Deep Dive into Moltbot

Lisa Ernst · 28.01.2026 · Artificial Intelligence · 8 min

The promise of a truly personal AI assistant, one that understands your unique workflows and operates directly on your devices, has long been a futuristic dream. Today, with the advent of open-source projects like Moltbot, that future is firmly within reach. This isn't about AI living in some distant cloud, but an intelligent assistant operating right where you communicate, offering genuine control and privacy.

Source: Own illustration

Quick Summary: Moltbot Installation & Key Features

• What it is: Moltbot (formerly Clawdbot) is an open-source, self-hosted personal AI assistant.
• Where it runs: macOS, Linux, Windows (via WSL2), Raspberry Pi, QNAP NAS, Hostinger VPS.
• Key Feature: Proactive communication, multi-agent routing, custom skills, digital workflow orchestration.
• Installation: Typically a single command line: curl -fsSL https://molt.bot/install.sh | bash.
• Requirements: Node.js v22+, API keys for LLM providers (Anthropic, OpenAI).
• Access: WhatsApp, Telegram, Discord, Slack, iMessage, Terminal User Interface (TUI), Web Dashboard.
• Security: Emphasizes sandboxing (Docker required), recommend installation on secondary devices due to access levels.
• Customization: Plugin-based architecture with a growing skills marketplace (ClawdHub).

Moltbot: An Overview of the Self-Hosted AI

Moltbot, previously known as Clawdbot, is an open-source, personal AI assistant designed to run directly on your own hardware. This includes operating systems like macOS, Linux, Windows (via WSL2), or even a Raspberry Pi. This emphasis on self-hosting provides users with full control over their data and ensures privacy. It integrates seamlessly with popular messaging platforms such as WhatsApp, Telegram, Discord, Slack, Signal, and iMessage.

Unlike traditional AI assistants that primarily react to prompts, Moltbot can proactively communicate and send messages. It leverages multi-agent routing, supports custom skills, and can orchestrate complex digital workflows. With over 30,000 GitHub stars, more than 8,900 Discord members, and over 130 contributors as of January 2026, Moltbot thrives on a strong, community-driven development model.

Source: cmswire.com

Anthropic, developer of the Claude AI model, requested the name change from Clawdbot to Moltbot to avoid confusion.

The project shifted its name from Clawdbot to Moltbot at the request of Anthropic, the developer of the Claude model, due to naming similarities. All previous documentation references to Clawdbot now refer to Moltbot, with corresponding changes to GitHub repositories, NPM packages, and official websites (Moltbot.bot). The clawdbot CLI command remains functional as a compatibility layer, though moltbot is now the recommended command.

Installation and Initial Setup

Setting up Moltbot typically begins with a single command line in your terminal. This streamlined approach makes initial deployment straightforward:

curl -fsSL https://molt.bot/install.sh | bash

For more detailed instructions, refer to the official Moltbot documentation. After the installation script runs, an onboarding assistant automatically launches (Moltbot Wizard). This assistant provides two main options:

• QuickStart: For those who prefer reasonable default settings.
• Advanced: For users who desire full configuration control.

During this process, a security warning appears, directing users to the documentation regarding sandboxing (Moltbot Gateway Security).

Prerequisites and API Keys

Moltbot requires Node.js version 22 or higher. To leverage its capabilities, you will need API keys for Large Language Model (LLM) providers such as Anthropic (Claude Pro/Max) or OpenAI. While API token costs can range from $10 to $150 per month depending on usage, a Claude Max or Pro subscription offers an alternative to pay-per-token models through a setup token.

Channel and Gateway Setup

The onboarding assistant also handles channel setup, allowing you to link messaging apps like WhatsApp by scanning a QR code from your phone’s linked devices settings (Moltbot WhatsApp Channel). Skills configuration can be deferred during the initial setup to streamline the process. The Gateway service, a crucial component for Moltbot’s operation, installs automatically (Moltbot Gateway).

Post-Installation and Access

Once installed, you can interact with Moltbot through various interfaces, catering to different preferences and use cases:

• Messaging Apps: WhatsApp, Telegram, Discord, Slack, iMessage.
• Terminal User Interface (TUI): Accessible by typing moltbot tui in your terminal.
• Web Dashboard: Available at http://127.0.0.1:18789/ (Moltbot Web Dashboard). Note that this is initially accessible only via localhost, requiring an SSH tunnel for remote access.

For managing the gateway, the command moltbot gateway --port 18789 --verbose can be used to start it in the foreground.

Deployment Options

Moltbot offers flexible deployment options beyond local machines, allowing for greater versatility:

• QNAP NAS: It can be installed on a QNAP NAS running Ubuntu Linux Station, with Ubuntu 24.04 being the recommended version. Access to the Ubuntu desktop for configuration is then possible via VNC.
• Cloud VPS: For cloud deployments, Moltbot can be installed on a Hostinger VPS using their one-click Docker template. This involves configuring environment variables like MOLTBOT_GATEWAY_TOKEN, ANTHROPIC_API_KEY, and OPENAI_API_KEY. The MOLTBOT_GATEWAY_TOKEN is particularly important for accessing the Moltbot web interface and should be stored securely. After a Hostinger installation, Moltbot becomes accessible at http://your-vps-ip:port.

Managing Moltbot and Its Capabilities

Moltbot stores conversational context in Markdown files under ~/clawd/, with USER.md containing user-specific facts and the memory/ directory holding long-term memories. This allows the AI to develop a more personalized understanding and response.

The assistant can proactively engage users through features like "Heartbeats" or Cron-Jobs. Cron-Jobs enable precise, scheduled tasks, such as daily morning briefings or automated reports.

Skills and Extensibility

Moltbot’s power lies in its extensible, plugin-based architecture, supported by a growing marketplace of skills known as ClawdHub. Skills are essentially folders containing a SKILL.md file and optional scripts or configurations, adhering to the AgentSkills specification, an open standard developed by Anthropic.

Moltbot comes bundled with 49 skills covering common workflows, including integrations with:

• Apple Ecosystem: Notes, Reminders.
• Google Workspace: Gmail, Calendar.
• Communication Platforms: Slack, iMessage.
• Smart Home Devices: Philips Hue.
• Development Tools: GitHub CLI.

Source: autocad123.vn

Moltbot integrates with communication platforms like Slack, enabling automated workflows and responses across your messaging channels.

Users can also create their own skills for specific tasks like image editing or Tesla control. The awesome-moltbot-skills repository features over 565 community-contributed skills. These skills can be installed via the ClawdHub CLI (npx clawdhub@latest install <skill-slug>) or manually.

Moltbot's Versatile Applications

Moltbot is designed as a bridge between AI models and your local files, applications, and automations. Here are some examples of its capabilities:

Security and Best Practices

Given Moltbot’s ability to run shell commands, execute scripts, and access sensitive data like password managers, security is paramount. Moltbot’s security documentation (Moltbot Security Documentation) emphasizes the importance of sandboxing. Sandboxing limits the AI agent’s access to the system, especially crucial when Moltbot operates on your primary device. Utilizing Docker is a prerequisite for sandboxing, requiring the sandbox image to be built from Moltbot’s source directory (Moltbot Docker Installation). For macOS users, granting "Full Disk Access" to the Terminal application in System Settings is necessary for remote access.

Security experts compare Moltbot to a butler with full access, underscoring the potential risks. Storing Moltbot’s "memories" as unencrypted text files creates a vulnerability if the device is compromised. There’s also a risk of Moltbot control servers becoming publicly accessible, potentially granting cyber attackers access. Therefore, installing Moltbot on secondary devices rather than primary machines containing sensitive data is strongly recommended. Users should thoroughly review and implement Moltbot’s security measures. The moltbot security audit command can help identify weak model tiers and secure sensitive paths.

Source: logos-marcas.com

Docker containerization is essential for Moltbot security, providing necessary sandboxing to protect your system during operation.

Moltbot supports local LLMs like Ollama, LM Studio, and Harbor, or any OpenAI-compatible API endpoint. GLM-4.7-Flash is noted as a strong local model for tool-calling capabilities. Moltbot complements IDE-focused tools like Claude Code or Cursor, allowing it to control and orchestrate them.

Conclusion

Moltbot represents a significant step towards a personalized, self-hosted AI assistant that seamlessly integrates into your digital life. Its open-source nature, community-driven development, and extensive customization options offer powerful capabilities for automating tasks and streamlining workflows. However, leveraging its full potential requires a thoughtful approach to installation and a firm understanding of its security implications to ensure a robust and private AI experience.

Sources