An AI assistant that lives on your server, not somebody else's
OpenClaw is an assistant you talk to in the apps you already use, and it does more than answer: it browses, works with files, runs commands and schedules tasks. Hosted on your own machine, its memory stays on your disk and its only outbound calls are to the model provider you choose.
- Talks on WhatsApp, Telegram, Slack
- Memory stays on your disk
- Your own model keys
- Root access from day one
We build the machine with OpenClaw already installed. You run the onboarding wizard once over SSH, and it is yours.
- WhatsApp
Where is the supplier report? Send me the short version.
- Telegram
Remind me at six about the call, and pull up their last invoice.
- Slack
Summarise this thread and post the decision in the channel.
What is OpenClaw?
An open-source AI assistant that runs on a server you own. You talk to it in the apps you already use, and it does the work rather than only describing it.
Not another chat window
It browses, edits files, runs commands and schedules work. A chatbot answers a question; this one goes and does the thing.
Reached where you already are
WhatsApp, Telegram, Slack, Discord and more. There is no separate app to open and nothing new for your team to learn.
It runs on your machine
The assistant, its memory and its files sit on your own disk. The only traffic leaving the server is the call to the model you chose.
Free and open source
The software itself costs nothing. What you pay us for is the server, the install, and keeping it running.
You bring the model
OpenClaw is the assistant, not the model. Add a key from your own provider, or point it at a model running locally.
Yours to change
Root access from the first hour. Put something else on the machine beside it, or take the whole thing elsewhere.
- WhatsApp
“Where is the supplier report? Send me the short version.”
Finds the file on your server, reads it, and replies in the same chat with the part you asked for.
- Telegram
“Remind me at six about the call, and pull up their last invoice.”
Sets the reminder as a scheduled job and fetches the invoice from your connected storage when it fires.
- Slack
“Summarise this thread and post the decision in the channel.”
Reads the conversation, writes the summary and posts it back itself — you do not copy anything across.
- Any chat
“Check the site every ten minutes and tell me if it goes down.”
Runs the check from your server on a schedule and messages you only when something has actually changed.
- Any chat
“Read this contract and tell me the notice period.”
Opens the document you sent, finds the clause and quotes it back with the wording around it.
- Any chat
“Go through last month's logs and tell me what kept failing.”
Hands the long job to a sub-agent that works in the background and reports when it has an answer.
These are the kinds of request people send it. What it can reach — your files, your calendar, your storage, your shell — is set by you when you run the onboarding wizard.
The machines it runs on
Every plan is a virtual machine of your own with root access, billed monthly in rupees. The application is installed before we hand it over.
OpenClaw runs on any of our Basic VPS machines. Choose by how much you will ask of it: a personal assistant needs far less than one running for a team with several sub-agents.
Basic 6 suits most people: enough headroom for continuous use, scheduled tasks and a few background jobs, without paying for capacity an assistant will not touch.
Prices are for the server. OpenClaw itself is free and open source, but the model behind it is not: you pay your AI provider per token, and an autonomous assistant can use a great many of them. Keep an eye on that bill as well as this one. Basic 4 has a one-time setup fee on a monthly contract, shown beside its price; there is none on yearly billing.
An assistant with hands, not just a chat window
The reason to self-host this one is that it acts. It reaches your files, your calendar and your shell, which is exactly why it should run somewhere you control.
Answers where you already talk
WhatsApp, Telegram, Discord, Slack, iMessage and Microsoft Teams.
Remembers across sessions
Preferences and context persist on your own disk rather than resetting each conversation.
Works on a schedule
Recurring tasks and proactive messages, without you asking each time.
Uses real tools
Browsing, file handling and shell commands, within whatever limits you set on the machine.
Delegates to sub-agents
Long jobs can be split off and run in the background while you carry on.
Keeps your data local
Nothing leaves the server except the calls to the model provider you chose.
The things it can reach
Named on the project's own integration list. What you connect is up to you, and each connection is yours to authorise.
- Telegram
- Discord
- Slack
- iMessage
- Microsoft Teams
- Anthropic Claude
- OpenAI GPT
- Google Gemini
- Ollama, local models
- Google Calendar
- Notion
- Todoist
- GitHub
- GitLab
- Home Assistant
- Philips Hue
- Sonos
- Spotify
- YouTube
- Dropbox
- Google Drive
- AWS S3
- Docker
- Webhooks
- CI and CD pipelines
The server is a fixed cost. The model is not.
OpenClaw is free to run, and your machine costs the same every month whatever it does. The model behind it charges per token, and an assistant that works on its own can spend a great deal without being asked twice. Set limits with your provider before you let it run unattended.
See VPS plansWorth knowing
You need a model key
OpenClaw does not ship with a model. You bring an API key from a provider, or point it at a local model you run yourself.
It is unmanaged
We build and support the server; the assistant and its configuration are yours, as with any root-access machine.
The interface is private
There is no public dashboard. You reach the control interface through an SSH tunnel, which is the right default for something with shell access.
Give it room
An assistant that browses, schedules and delegates uses more than one that answers questions. Start at a size with headroom.
Three things happen, and only one is yours
- 1
We build the machine
You choose a size and a location; we build the server with OpenClaw already installed and send you the credentials.
- 2
You run the onboarding wizard
Over SSH, once. Add your model provider key, connect the messaging apps you want, and set your preferences.
- 3
You talk to it
From the terminal, or through the control interface over an SSH tunnel. After that it is running, and it remembers.
We support the server, the network and the installation. What the assistant does after that is yours to configure, as with any unmanaged machine.
Other things we host and build
OpenClaw Hosting, answered
What is OpenClaw?
An open-source AI assistant that runs on your own server and talks to you through the messaging apps you already use. Unlike a hosted chatbot it can browse, handle files, run commands and schedule work.
What is OpenClaw?
An open-source AI assistant that runs on your own server and talks to you through the messaging apps you already use. Unlike a hosted chatbot it can browse, handle files, run commands and schedule work.
Give your assistant a machine of its own
Pick a size and a location, and we will build the server with OpenClaw installed and ready for its first run.