The Complete Guide to Clawdbot/Moltbot

Your Self-Hosted Personal AI Assistant

Moltbot (previously called Clawdbot) is an open-source AI assistant you can run on your own hardware. Think of it as building your own Jarvis: an AI that doesn't just answer questions, but can actually do things for you: manage your calendar, send emails, automate tasks, and more.

Created by Peter Steinberger, it exploded in popularity in January 2026, gaining nearly 30,000 GitHub stars shortly after launch. The rename to "Moltbot" happened on January 27, 2026, after Anthropic raised trademark concerns; the mascot evolved from "Clawd" to "Molty," a nod to how lobsters molt to grow.

Table of Contents

• Your Self-Hosted Personal AI Assistant
• Table of Contents
• What Makes Moltbot Different
• Hardware Requirements
• Minimum Specs (Basic Operation)
• For 24/7 Always-On Operation
• If You Want to Run Local AI Models
• The Critical Security Rule
• The Golden Rule: Use Separate Accounts for Everything
• Best Practices
• Choosing Your AI Brain
• Premium Choice: Claude Opus 4.5
• Budget-Friendly: MiniMax 2.1
• Other Options
• Pro Tip: Mix and Match
• Cost Reality Check
• Installation Walkthrough
• The Easy Way: One-Liner Installation
• The Manual Way: For More Control
• After Installation: Adding Skills
• Connecting Your Messaging App
• Verify Everything Works
• The Training Period
• The First 2-3 Weeks Are Critical
• How to Train It
• Make It Proactive
• Where Moltbot Actually Shines (And Where It Doesn't)
• ✅ The Sweet Spot: Where Moltbot Excels
• ⚠️ The Honest Middle Ground: It Works, But...
• ❌ Where Moltbot Is Probably Not the Right Choice
• The Ideal Moltbot User
• Real Talk: The Investment vs. Payoff
• What Can You Actually Do With It?
• Personal Productivity
• Research and Analysis
• Automation
• Content and Trends
• Agentic Workflows
• Development and Creative Work
• Business Operations
• Building Your Second Brain
• Setting It Up
• Feeding Your Second Brain
• Querying Your Knowledge
• Maintenance
• Limitations and Risks
• Functional Limitations
• Security Risks
• Privacy and Data Concerns
• Reliability Issues
• Learning Curve
• Community Risks
• Final Thoughts

What Makes Moltbot Different

Most AI assistants live in the cloud and can only talk. Moltbot is different in two fundamental ways:

1. It lives on your hardware. Your conversations, data, and automations stay on your own machine or server. No corporate cloud required.
2. It has "arms." This is the key concept. Moltbot isn't just a chatbot; it can execute real actions:
• Control your browser
• Read and write files
• Make API calls to services
• Send messages on your behalf
• Run scheduled tasks

This makes it extraordinarily powerful, but it also means you need to treat it with respect. More on that shortly.

Hardware Requirements

Moltbot is designed to be lightweight and flexible. Here's what you actually need:

Minimum Specs (Basic Operation)

For 24/7 Always-On Operation

If You Want to Run Local AI Models

Running AI models locally (instead of using cloud APIs) requires more muscle:

• 8-16GB RAM
• A dedicated GPU (NVIDIA with CUDA support is ideal)

The Critical Security Rule

The Golden Rule: Use Separate Accounts for Everything

Never connect Moltbot to your primary accounts. Create dedicated, isolated accounts for:

• Email (new Gmail or similar)
• Calendar
• GitHub
• Any other service it will access

Imagine Moltbot accessing your real email and being tricked into sending messages to your contacts. Or modifying files on your main computer. By isolating it to dedicated accounts, any potential damage stays contained.

Best Practices

1. Create burner accounts during setup
2. Configure permissions narrowly (read-only wherever possible)
3. Enable two-factor authentication on all connected accounts
4. Don't run it on your primary computer: use a virtual private server (VPS) like AWS or dedicated machine
5. Use Docker to containerize the installation for additional isolation
6. Never expose ports publicly if running on a VPS

The community consensus is clear: treat Moltbot as "an AI with arms in a sandbox." Give it access to what it needs, nothing more.

Choosing Your AI Brain

Moltbot is essentially a gateway; it routes your requests to an AI model that does the actual thinking. You can choose from several options:

Premium Choice: Claude Opus 4.5

Anthropic's flagship model offers:

• Superior reasoning capabilities
• Excellent long-context handling
• Better resistance to prompt injections
• Ideal for complex tasks like autonomous research or code work

Cost: $100-200/month for Pro/Max tiers

Budget-Friendly: MiniMax 2.1

A strong option for cost-conscious setups:

• Can run locally or via API
• Roughly $0.01-0.05 per 1,000 tokens
• Works well on consumer hardware
• Great for everyday automation tasks

Other Options

• Ollama for local model discovery (runs models like Llama 3)
• OpenAI GPT-4o if you prefer that ecosystem
• Grok or any OpenAI-compatible endpoint
• ElevenLabs or Edge TTS for voice features

Pro Tip: Mix and Match

Many users configure multiple models for different purposes. For example, Claude for planning and complex reasoning, MiniMax for routine execution tasks. Moltbot's routing system makes this straightforward.

Cost Reality Check

One user reported burning through 180 million tokens in a single week. API costs can add up quickly with heavy usage. Local models eliminate this concern but require beefier hardware.

Installation Walkthrough

The Easy Way: One-Liner Installation

The simplest way to get started is the one-liner that handles everything for you, including installing Node.js if you don't have it.

macOS / Linux:

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

Windows (PowerShell):

iwr -useb https://molt.bot/install.ps1 | iex

That's it. The script installs Node.js (if needed) and sets up Moltbot automatically. It works on macOS, Windows, and Linux.

After installation completes, the onboarding wizard will guide you through:

• Entering your AI provider API keys
• Connecting your messaging app
• Basic configuration

The Manual Way: For More Control

If you prefer to handle things step-by-step (or the one-liner doesn't work for your setup), here's the detailed approach:

Prerequisites:

1. Install Node.js (version 22+) from nodejs.org
2. Create API keys for your chosen AI provider(s)
3. Set up a bot token for your messaging app (e.g., Telegram via BotFather)
4. If on Windows, WSL2 is recommended

Step 1: Install via npm

npx moltbot@latest

Or for a global installation:

npm install -g moltbot@latest

Step 2: Run the Onboarding

moltbot onboard --install-daemon

This configures Moltbot to run continuously in the background and walks you through initial setup.

Step 3: Configure Your AI Provider

During onboarding, you'll enter your API keys. These get stored in ~/.clawd/clawdbot.json (yes, the folder still uses the old name during the transition period).

After Installation: Adding Skills

"Skills" are plugins that give Moltbot specific capabilities: browser control, scheduled tasks, file manipulation, and more.

Browse available skills at clawdhub.com, or install via command line:

moltbot skills install <skill-name>

Most bundled skills are free; some community or managed skills may have costs.

Connecting Your Messaging App

Choose your preferred way to communicate with Moltbot:

Configure your chosen channel in the Control UI or config file.

Verify Everything Works

Access the Control UI at localhost:31415 to monitor sessions and configurations. Try a simple test:

"Remind me in 5 minutes to check the oven."

If that works, you're ready for the training period.

The Training Period

The First 2-3 Weeks Are Critical

This is where Moltbot transforms from a generic assistant into your assistant. The system uses vector memory (introduced in v2026.1.12) to store and recall information about you persistently.

How to Train It

1. Start with a "Getting to Know You" Conversation

One of the most effective ways to kick off your training period is to let Moltbot interview you. Instead of trying to think of everything it needs to know, flip the script:

"Hi Moltbot, let's get to know each other and build a good working relationship. I'm going to tell you about myself, then please ask me any questions you want!"

This approach works surprisingly well. The AI will ask about your work, preferences, schedule, communication style, and goals: often surfacing things you wouldn't have thought to mention. It's a more natural way to build context than writing a giant information dump.

2. Daily Briefs: Start each day by telling Moltbot your plans, priorities, and any context it should know.
3. Share Your Preferences: Be explicit about how you like things done.
• "I prefer concise responses unless I ask for detail."
• "When scheduling meetings, always leave 15-minute buffers."
• "My work hours are 9 AM to 6 PM, Monday through Friday."
4. Provide Context Dumps: Share documents about your routines, goals, projects, and preferences. The more context, the better it can anticipate your needs.
5. Correct Mistakes: When it gets something wrong, explain why. It learns from these corrections.
6. Enable Tools Gradually: Don't give it access to everything on day one. Add capabilities as you build trust and understand its behavior.

What to Expect

Week 1: Getting acquainted. Lots of explicit instructions, occasional corrections, building basic context.

Week 2: Improved accuracy. Starts anticipating needs, remembers preferences, fewer corrections needed.

Week 3: Autonomous operation. Handles routines without prompting, generates useful proactive suggestions, feels genuinely personalized.

Make It Proactive

Use "hooks" to trigger proactive behavior:

• Email arrives → Moltbot summarizes and suggests actions
• Calendar event approaching → Moltbot prepares briefing materials
• Specific time of day → Moltbot delivers morning news digest

Where Moltbot Actually Shines (And Where It Doesn't)

Let's be honest about what Moltbot is genuinely good for, and where you might be better served by other solutions. The hype is real, but so is the reality check.

✅ The Sweet Spot: Where Moltbot Excels

1. You value privacy and data sovereignty above convenience.

If the idea of your personal data, conversations, and automations living on someone else's servers genuinely bothers you, Moltbot is built for you. Everything stays on your hardware. No corporate AI company is training on your private life.

2. You have repetitive workflows that cross multiple tools. Moltbot shines when you need to connect things that don't naturally connect. "When I get an email from this client, check my calendar, draft a response, and add a task to my project tracker". This kind of multi-step, multi-tool automation is where the "AI with arms" concept pays off.

3. You're technical enough to troubleshoot (or willing to learn). Moltbot rewards users who can read error logs, tweak configurations, and debug issues. If that sounds appealing rather than exhausting, you're the target audience.

4. You want a genuinely personalized assistant, not a generic one. After the training period, Moltbot knows your preferences, your schedule, your communication style. It's not trying to be helpful to everyone, it's trying to be helpful to you specifically. This deep personalization is hard to replicate with cloud services that reset context constantly.

5. You're building a personal knowledge system. The "second brain" use case is compelling. If you're someone who takes notes, collects information, and wants an AI that can actually retrieve and synthesize your accumulated knowledge, Moltbot handles this well.

⚠️ The Honest Middle Ground: It Works, But...

1. Simple task automation. Yes, Moltbot can set reminders and manage your calendar. But if that's all you need, you're using a chainsaw to cut butter. Native phone assistants or simple apps like Todoist handle basic tasks with zero setup.
2. Occasional research tasks. Moltbot can research topics and compile information. But for one-off research, you might get faster results just using Claude, ChatGPT, or Perplexity directly. Moltbot's value emerges when research is recurring, personalized, or needs to integrate with your other systems.
3. Team collaboration. Moltbot is fundamentally a personal assistant. While it can operate in group chats and handle some team coordination, it wasn't designed as a team tool. For collaborative workflows, purpose-built tools (Slack integrations, Notion AI, etc.) might serve you better.

❌ Where Moltbot Is Probably Not the Right Choice

1. You want plug-and-play simplicity. If you're looking for something that "just works" out of the box, Moltbot will frustrate you. The setup, training period, and ongoing maintenance require real investment. There's no shame in preferring consumer-friendly alternatives.
2. You're not comfortable with security trade-offs. Running an autonomous AI with system access is inherently risky. If the security considerations in this guide made you nervous rather than intrigued, trust that instinct. A compromised Moltbot installation can do real damage.
3. Your workflows are already well-served by existing tools. If Zapier, IFTTT, or native app integrations already handle your automation needs, Moltbot adds complexity without clear benefit. It's most valuable when existing tools can't do what you need.
4. You need enterprise-grade reliability. Moltbot is an open-source project, not an enterprise product. It can hang, hallucinate, or behave unexpectedly. For mission-critical business processes, purpose-built enterprise automation tools offer guarantees Moltbot cannot.
5. You're primarily mobile. There's no native mobile app. You interact via messaging apps, which works, but the experience is fundamentally designed around desktop/server operation. If you're rarely at a computer, the value proposition diminishes.

The Ideal Moltbot User

If we're being completely honest, Moltbot is ideal for a specific profile:

A technically comfortable individual who values privacy, has complex personal or professional workflows spanning multiple tools, is willing to invest 2-3 weeks in training, and finds the idea of a deeply personalized AI assistant genuinely exciting rather than abstractly interesting.

If that sounds like you, Moltbot can be transformative. If it doesn't, there's no judgment, plenty of excellent alternatives exist for different needs.

Real Talk: The Investment vs. Payoff

What Can You Actually Do With It?

Personal Productivity

• Morning briefings (weather, calendar, news relevant to you)
• Automatic calendar time-blocking
• Weekly reviews compiled from your notes and transcripts
• Task prioritization with urgency scoring

Research and Analysis

• Break projects into actionable tasks
• Research people before meetings (compile briefing documents)
• Analyze documents and extract key insights
• Monitor topics across multiple sources

Automation

• Invoice processing and expense tracking
• Smart home control
• Portfolio monitoring and optimization
• Handling bureaucratic tasks (yes, even phone calls in some configurations)

Content and Trends

• Monitor social platforms for opportunities
• RSS feed summaries
• Trend analysis in your areas of interest

Agentic Workflows

• Spawn sub-agents for brainstorming
• Manage competing priorities and surface conflicts
• Family coordination (school schedules, appointments, reminders)

Development and Creative Work

• Code refactoring and review
• Browser automation for testing
• Dynamic UI generation via Canvas feature

Business Operations

• Lead generation
• Customer support in group chats
• Integration with Gmail, Slack, and other business tools

Building Your Second Brain

Moltbot can serve as a persistent knowledge management system, a "second brain" that organizes and retrieves your information.

Setting It Up

Create a dedicated folder structure in ~/clawd/:

~/clawd/
├── knowledge/      # General reference material
├── notes/          # Meeting notes, ideas, observations
├── projects/       # Active project documentation
├── habits/         # Routines and personal systems
└── preferences/    # How you like things done

Feeding Your Second Brain

• Upload PDFs, documents, Notion exports
• Connect email for automatic ingestion (via Gmail Pub/Sub)
• Use skills to index content from various sources

Querying Your Knowledge

The vector search capability lets you ask questions in natural language:

"What did we decide about the Q3 marketing strategy?" "Find my notes from the client meeting last Tuesday." "What are my key habits for morning productivity?"

Maintenance

Set up cron jobs for automatic backups and use hooks to keep information current as things change.

Limitations and Risks

Functional Limitations

• Permission dependencies: Some features require system-level access (macOS TCC for camera, for example)
• Group chat noise: Needs mention-gating to avoid responding to every message
• Local model speed: Running AI locally is slower than cloud APIs
• No native mobile app: Uses nodes for iOS/Android features like camera access
• Sandboxing is optional: The default setup doesn't isolate the main session

Security Risks

1. Prompt Injection: This is the big one. Malicious content from any source (DMs, webpages, documents) can potentially hijack Moltbot's behavior. The sandbox approach described earlier is your primary defense.
2. Public Exposure: Running on a publicly accessible VPS without proper security invites brute-force attacks and worse.
3. Fake Extensions: There have been instances of malware disguised as Moltbot-related tools (including fake VS Code extensions containing remote access trojans).

Privacy and Data Concerns

• Sensitive information can leak if the system isn't properly isolated
• Autonomous actions might unintentionally modify or delete files
• Be thoughtful about what you allow it to access

Reliability Issues

• Broken command-line tools can cause hangs
• API costs can escalate quickly ($50-200/month isn't unusual for heavy users)
• AI models can hallucinate or ignore safety policies

Learning Curve

Moltbot is not a plug-and-play consumer product. It requires:

• Comfort with command-line tools
• Understanding of security concepts
• Willingness to monitor and adjust during the training period
• Time investment to configure properly

Community Risks

The rapid popularity has attracted scammers. Fake cryptocurrency tokens claiming association with the project have crashed 90%+ in value. Always verify official sources.

Final Thoughts

Moltbot represents a fascinating frontier: genuinely autonomous AI assistants that respect your privacy and run on your infrastructure. The power it offers is real, but so are the responsibilities that come with it.