
Remember when Jarvis felt like pure science fiction? Tony Stark talks to his AI, it manages his schedule, controls his home, runs research in the background, and proactively alerts him to important things.
That future is here. And it costs about $5 a month.
Clawdbot is an open-source AI assistant that lives in your messaging apps. You text it like a friend. It texts you back. It remembers every conversation you've ever had. It can reach out to you first when something important happens. And it can actually do things on your computer.
This guide covers everything you need to know to set up your own.
Clawdbot is an open-source AI framework created by Peter Steinberger, a developer from Vienna who came back from retirement to build it. The project has an MIT license, a space lobster mascot named Clawd, and an endorsement from Andrej Karpathy.
Here's the core idea: instead of visiting an AI on a website, the AI lives in apps you already use. WhatsApp. Telegram. iMessage. Discord. Slack. You message it from your phone, laptop, or tablet. Same conversation, same context, same memory across all devices.
Three things make it different from ChatGPT or Claude's web interface:
It remembers everything. Ask it about something you mentioned two weeks ago. It knows. It builds context over time and gets better at helping you.
It messages you first. Normal AI waits for you to open it. Clawdbot can reach out proactively. Morning briefings with your calendar and important emails. Alerts when something you're tracking changes. Reminders you set once that run forever.
It does real tasks. Not just answering questions. Actually doing things. Sending emails. Managing files. Running programs. Controlling your browser. One person rebuilt their entire website while watching TV in bed. Never opened a laptop. Just texted their assistant what to do.
If you're using AI tools like Cursor or Claude Code to build software, Clawdbot adds a layer on top of everything.
It can control coding tools like Claude Code. You can tell it to research a trending topic on the web, then build an app based on what it finds. You can ask it to "build something cool" before you go to sleep and wake up to progress.
Think of it as vibe coding on autopilot. You're not sitting at your desk prompting an AI in a code editor. You're texting your assistant from anywhere, and it's executing in the background.
Some developers have it manage their entire development workflow. Reviewing repositories. Running tests. Opening pull requests. Managing tasks through a Kanban board it built itself.
Here's where people overcomplicate things.
Search for Clawdbot setups and you'll see photos of Mac Mini farms. Multiple machines stacked on desks. Raspberry Pi clusters. People treating this like they need a data center.
You don't.
Clawdbot runs perfectly fine on a $5/month cloud server. That's less than a coffee.
When a cheap VPS makes sense (most people):
When a Mac Mini makes sense (advanced users):
For most people starting out, a VPS is the practical path. You can always upgrade later.
Let's be clear about what this costs:
| Component | Cost |
|---|---|
| Clawdbot software | Free (open source) |
| Server (Hetzner VPS) | ~$5/month |
| AI model (Claude Pro) | $20/month |
| AI model (Claude Max) | $100-200/month |
| Total | $25-150/month |
If you already have a Claude Pro or Max subscription, you can use that instead of paying for API credits separately. The setup wizard supports this.
Compare that to hiring an actual assistant. Or the "AI consultants" charging thousands to set up basic chatbots.
This takes about 15-20 minutes. You'll need to be comfortable copy-pasting commands into a terminal, but the process is straightforward.
Step 1: Sign up for Hetzner
Go to hetzner.com/cloud and create an account. Hetzner offers cost-optimized servers in Germany that work well for this.
Step 2: Create a new project
In the Cloud Console, click "+ New Project" and name it something like "clawdbot".
Step 3: Create your server
Click into your project, then "Create Server". Choose these settings:
Step 4: Add your SSH key
If you don't have an SSH key, open Terminal on your Mac and run:
ssh-keygen -t ed25519 -C "your@email.com"
Press Enter to accept defaults. Then display your public key:
cat ~/.ssh/id_ed25519.pub
Copy the output and paste it into Hetzner's SSH key field.
Step 5: Create and note the IP
Name your server "clawdbot" and click "Create & Buy Now". Wait about 30 seconds for it to spin up, then copy the IPv4 address.
Step 6: Connect to your server
ssh root@YOUR_SERVER_IP
Replace YOUR_SERVER_IP with the IP you copied. Type "yes" when asked about the fingerprint.
Step 7: Update the system
apt update && apt upgrade -y
Step 8: Install Node.js 22
curl -fsSL https://deb.nodesource.com/setup_22.x | bash -
apt install -y nodejs
Verify with node -v. Should show v22.x.x.
Step 9: Install Clawdbot
npm install -g clawdbot@latest
Step 10: Run the onboarding wizard
clawdbot onboard --install-daemon
This wizard walks you through everything: model authentication, workspace setup, and channel configuration.
Telegram is the easiest channel to set up. You can add others later.
Step 11: Create a Telegram bot
/newbotStep 12: Get your Telegram user ID
Search for @userinfobot on Telegram and send it any message. It replies with your numeric user ID. Copy this number.
Step 13: Enter credentials in wizard
When the Clawdbot wizard asks:
Step 14: Get your API key
Go to console.anthropic.com and create an API key. It starts with sk-ant-...
If you have Claude Pro or Max, you can use your subscription instead of API credits. The wizard will offer this option.
Step 15: Enter in wizard
When the wizard asks for Anthropic auth, choose your method and paste the key or token.
Step 16: Complete the wizard
Follow the remaining prompts, accepting defaults for most options. The wizard installs a background daemon so Clawdbot stays running.
Step 17: Check status
clawdbot status
Should show the gateway as running.
Step 18: Test it
Open Telegram, find your bot, and send it a message. You should get a response.
Congratulations. You have a Jarvis.
Your Clawdbot workspace (usually ~/clawd) contains files that define your assistant's personality and knowledge.
SOUL.md or IDENTITY.md
This defines your assistant's personality. Example:
# Assistant
You are a helpful assistant. You're:
- Direct and concise
- Proactive with suggestions
- Knowledgeable about my projects and goals
Your communication style:
- Skip pleasantries unless I initiate
- Always provide actionable next steps
- Challenge me when I'm making excuses
USER.md
Information about you that helps the assistant be more useful:
# User Profile
Name: [Your Name]
Timezone: PST
Work schedule: 9am-6pm weekdays
Current projects:
- [List your active projects]
Preferences:
- Morning briefs at 7am
- Weekly reviews on Sundays
MEMORY.md
Persistent knowledge that grows over time. The assistant updates this as it learns about you.
Skills
Skills are plugins that extend what your assistant can do. Browse and install them from ClawdHub (clawdhub.com). Popular ones include web search, browser control, GitHub integration, and calendar management.
You can also ask your assistant to create custom skills. "Create a skill that monitors my competitor's pricing page and alerts me when prices change." It generates the code, you review it, and it's added to your skills.
Server management:
| Command | What it does |
|---|---|
clawdbot status | Check if everything is working |
clawdbot logs --follow | View live logs |
clawdbot gateway restart | Restart the bot |
clawdbot health | Run health checks |
Chat commands (send in Telegram):
| Command | What it does |
|---|---|
/new | Start a fresh conversation |
/model | Switch AI models |
/compact | Compress long conversations |
stop | Cancel a running task |
Based on what people are building:
Morning briefings. Wake up to a summary of your important emails, calendar for the day, and tasks you need to handle. Delivered to your phone before you get out of bed.
Email management. "Unsubscribe me from all these newsletters." It logs into your email, finds the junk, handles it.
Research assistant. "Find me the 5 best restaurants near my hotel in Tokyo." It searches, compares, and gives you options in a text thread.
Vibe coding workflows. "Research trending apps this week and prototype something interesting." It investigates, creates a project outline, and starts building.
Task automation. "Every Friday at 5pm, send me a summary of what I accomplished this week." Set it once, it runs forever.
Health tracking. Connect it to services like WHOOP and get daily fitness summaries automatically.
Content pipelines. Record voice notes during walks. The assistant transcribes them, extracts key ideas, drafts posts, and queues them for your review.
Second brain. Have it organize your ideas, thoughts, and research into a structured file system. Some users have replaced Notion entirely.
This is important. Clawdbot has no guardrails by design. That's what makes it powerful, but it also means you need to think carefully about what you give it access to.
Autonomous access means autonomous mistakes. If your assistant can send emails, it can send the wrong email. If it can manage files, it can delete the wrong file. If it misinterprets an instruction or hallucinates, it will act on that misinterpretation. Tread cautiously with what access you grant.
Start limited, expand as you build trust. Don't give it access to everything on day one. Start with messaging and research. Add email access once you trust it. Add file management later.
Use the pairing system. When someone first messages your bot, they need a pairing code that you approve. Don't disable this for DMs.
Allowlist group access. Don't let your assistant join random group chats. Use allowlist mode to control which groups it can participate in.
Keep sandboxing enabled. Unless you specifically need host access for a task, keep the sandbox on. It limits potential damage from unexpected behavior.
Review custom skills before installing. Even if the assistant wrote the skill itself, review the code. Make sure you understand what it does before enabling it.
Secure your API keys. Store them in environment variables or secure config files. Never commit them to git repositories.
Back up your workspace regularly. Your ~/clawd directory contains all your assistant's memory and configuration. Back it up.
Yes if:
Maybe wait if:
The honest take: Clawdbot is early and moving fast. Bugs get fixed quickly. Features ship weekly. The Discord community has dozens of contributions per day. It's not polished enterprise software, but that's also why people using it now are building workflows that weren't possible before.
Here's the practical path:
Day 1: Install and connect Telegram. Have conversations to build context and memory.
Day 2: Customize SOUL.md and USER.md. Tell it about yourself, your projects, your preferences.
Day 3: Install the web search skill. Start using it for research tasks.
Day 4: Set up morning briefs. Connect your calendar if you want.
Day 5: Explore vibe coding workflows. Try having it research and prototype something.
Day 6: Add more channels if needed (WhatsApp, Discord, etc.).
Day 7: Review what's working. Refine. Add custom skills for your specific needs.
The key is starting simple. One channel, basic functionality. Then expand as you see what's possible.
Most people try to configure everything at once, get overwhelmed, and quit. The ones who succeed ship a basic version, use it daily, and incrementally add capabilities.
Once you have the basics running:
The assistant should come to you, not the other way around.
That's Clawdbot.