OpenClaw

Personal AI assistant on your own GPU infrastructureAn open-source assistant that lives where you do: WhatsApp, Slack, Discord, Telegram, iMessage. Runs locally on your GPU — your data never leaves your infrastructure.
374kTypeScriptMIT

Run OpenClaw on a QuData GPU

Personal AI assistant on a rented GPU. No cloud APIs, no token quotas.
Open Source MITRuns on your GPU22+ channels included
npm install -g openclaw@latest

Capabilities

What ships out of the box — no add-ons, no paid modules.
Any messaging channelOne process, dozens of platforms: WhatsApp, Telegram, Slack, Discord, iMessage, Teams, Matrix and more.
Live CanvasThe agent draws charts, forms and kanbans on a shared canvas you can both see and edit.
Isolated workspacesChannels, accounts and tools split into separate agents, each with its own memory and scope.
Secure sandboxTasks run in isolated environments with SSH and remote OpenShell backends available.
OpenAI-compatible APIPlug in local LLMs and any provider through a single protocol — no integration rewrites.
Extensible skillsBundled, managed and workspace skills plus the ClawHub registry for reusable playbooks.

Use cases

Examples already running on QuData infrastructure.
Customer supportPicks up messages across channels, answers from your knowledge base, escalates only the hard cases.
Internal team copilotHelps with code, reviews, docs and tickets right inside Slack or Teams.
Infrastructure operationsReacts to alerts, runs commands in a sandbox and logs every action — no on-call human needed.
Collaborative data workLive Canvas builds dashboards and forms inline during the conversation — no window switching.

How to run it locally

Step-by-step install with no external service dependencies.
  1. 1
    Prepare the environment: Node.js 22.19+ and npm
  2. 2
    Install OpenClaw globally
    npm install -g openclaw@latest
  3. 3
    Run the onboarding wizard
    openclaw onboard --install-daemon
  4. 4
    Connect a model and channels
  5. 5
    Open the Control UI and Live Canvas
    openclaw

Run OpenClaw on a rented GPU

QuData GPUs are available hourly or monthly. No overcommit, no hidden limits, no vendor lock-in.
FAQ

Frequently asked about OpenClaw

Where does my data go?

Nowhere. OpenClaw talks to a local LLM on a QuData-rented GPU. Messages, documents and keys never leave your infrastructure.

What GPU do I need to run it?

A single NVIDIA RTX 4090 or L40S is enough for light scenarios. For production loads pick an A100 or H100 — performance scales by an order of magnitude.

Why is this cheaper than cloud APIs?

You pay for an hour of GPU, not for every token. Under intensive load OpenClaw on a QuData A100 is 3–5x cheaper than an equivalent cloud API.

Can I extend the agent with my own logic?

Yes. OpenClaw supports workspace skills, the ClawHub registry and MCP — you can add tools without touching the core.