Clawdbot: a real AI assistant you can set up in 30 minutes
Most AI tools answer questions.
Clawdbot does the work.
It runs all day on a server. You talk to it through Telegram or WhatsApp. And instead of just replying, it actually carries out tasks. The boring ones. The ones that pile up.
Think inbox cleanup. Calendar juggling. Research. Follow-ups. Content drafts. Quiet background work that normally eats your time.
Once it’s running, you stop asking “can you help?” and start saying “do this.”
What Clawdbot actually is
Clawdbot is an open-source AI assistant that runs 24/7 on its own server. You don’t open a dashboard. You don’t log into another app.
You message it.
That’s it.
Behind the scenes, it can connect to things you already use like Gmail, Google Drive, GitHub, and calendars. You give it a task. It breaks the task into steps and runs them.
For example, you can say: Analyze my site, write a blog post, update the metadata, then draft a LinkedIn post.
It does all of that in sequence.
Another setup might monitor stock prices, check you into flights, or send alerts when something needs attention. You can also talk to it using voice notes if typing feels like friction.
Most AI tools stop at answers.
Clawdbot is built for execution.
Why the setup of Clawdbot is easier than people expect
People assume this means terminals, server headaches, and wrestling with APIs for hours.
Honestly, it doesn’t.
Most of the setup is copy-paste followed by a wizard that asks clear questions. If you ever get stuck, you can screenshot the step and ask ChatGPT what to click. That works more often than it should.
If you can follow basic instructions, you can set this up.
The full setup, start to finish
First, give it a place to live
Clawdbot needs to run somewhere all the time. A free AWS server works fine.
You go to aws.amazon.com and create an account. Open EC2. Launch a new instance. Name it whatever you want. Choose Ubuntu. Filter for free-tier eligible and pick the 8GB option.
Launch it.
Click the instance ID, hit Connect, then Connect again.
You’re now in a terminal. That’s the scariest part, and it’s already over.
Install Clawdbot
You paste one line.
curl -fsSL https://clawd.bot/install.sh | bash
That’s it.
Wait a couple of minutes. No config files. No dependencies to hunt down. When it finishes, the wizard starts automatically.
Run the wizard
Choose Quick Start.
Select Anthropic as the provider. When asked, paste your token. The wizard will tell you exactly how to get it by running a command on your local machine.
Choose Opus 4.5 as the model.
Select Telegram as your communication channel.
You’re nearly done.
Create your Telegram bot
Open Telegram and search for @BotFather. Send /newbot. Name it. Copy the token it gives you and paste it into the wizard.
Then search for @UserIDBot. Grab your user ID and paste that in too.
This step matters. It locks the bot so only you can talk to it.
Give the assistant an identity
Clawdbot will now message you on Telegram and ask a few simple questions.
What should it call you? What should you call it? What’s its purpose? What timezone are you in?
Answer honestly. This shapes how it behaves.
Once you’re done, the assistant is live.
Quick things to test right away
Start small. You want proof it’s useful before you get fancy.
Ask it to check your last ten emails and tell you which ones actually need a reply.
Ask it to research a company you’re curious about and summarize what they do.
Ask it to remind you tomorrow morning about something you’ve been avoiding.
If those work, everything else will too.
How much does Clawdbot cost
You pay for the AI model.
That’s about $20 a month for Claude.
The AWS server can stay on the free tier.
A human assistant costs hundreds or thousands per month and clocks out at night. Clawdbot runs nonstop and doesn’t forget things.
What to do after setup
Once it’s running, you can slowly make it smarter.
Add web search using Brave. Go to brave.com/search/api, grab a free API key, and tell your bot to set it up.
Connect your tools over time. GitHub. Google Drive. Gmail. Your calendar. No rush.
Then start giving it real tasks.
Research a company and turn it into a one-pager. Draft a LinkedIn post about a topic you care about. Check your calendar and find time for a call next Tuesday. Summarize an article and write a thank-you email to the author.
If something breaks, paste the error into the chat and say “fix this.” It often does.
You don’t have to type
Telegram supports voice notes.
You can talk to Clawdbot while walking, driving, or thinking out loud. It listens, processes, and responds like a teammate who never interrupts.
Some people even run multiple agents in Slack, firing off tasks in parallel. One researches. One writes. One monitors.
That part is optional. But it’s there.
Where to learn more
The docs live at clawd.bot
They’re straightforward. No fluff. Just how things work.
If you’ve been waiting for an AI assistant that actually does things instead of talking about them, this is one of the few that delivers.
Want to build your own agent or customize one for your workflow? Reach out and we’ll help you set it up the right way.