IHow it works

From a job description to a teammate, in four steps.

Pick a role, connect the apps it should use, decide what needs your approval, and turn it on. We host it, run it, and keep it up to date.

Hire an AI agentSee the four steps
Fig. 1. Four orbs on a gold band, waxing from dark to full: an agent from the day it is hired to the day it goes live. Painted in the browser; move the pointer over it.

IIThe four steps

Hiring, without the onboarding.

Step I. One agent, chosen from five roles, waits with its lit side toward the role it was given.
  1. Step I

    Pick a role

    Start from a job, not a blank canvas. Each role arrives knowing its work: what to answer, what to draft, when to ask.

    • Support agent
    • Sales and outreach
    • Email assistant
    • Executive assistant
    • Research assistant
  2. Step II

    Connect its apps

    Give it the places it should work: an inbox, a Slack workspace, a Telegram bot, Google Workspace. It works there the way a new hire would, under its own name.

    • Email
    • Slack
    • Telegram
    • WhatsApp
    • Discord
    • Google Workspace
  3. Step III

    Set the rules

    Decide, action by action, what runs on its own, what waits for your approval, and what goes straight to a person.

    • Runs on its own
    • Waits for your approval
    • Hands off to a person
  4. Step IV

    Turn it on

    We host it and run it. The agent reports to your dashboard, where every conversation, approval and handoff is kept.

    • Hosted for you
    • Every action logged
    • Pause or change it any time

IIIThe rules

You decide what it may do alone.

Every action gets one of three rules. A typical support agent, set up on its first day, looks like this.

When the agentRule
Answers a question your help centre already coversRuns on its own
Drafts a follow-up email to a lead who went quietRuns on its own
Issues a refund, or promises a delivery dateWaits for your approval
Moves a meeting on your calendarWaits for your approval
An upset customer asks for a managerHands off to a person
Fig. 3. An example rule set. A full orb runs on its own, a half orb waits for approval, a dark orb goes to a person.

IVWhat we run for you

You never see a server.

Hosting
Each agent runs in its own workspace on servers we manage. We install it, update it and restart it when it needs restarting.
Your model, your key
Anthropic, OpenAI, Gemini, OpenRouter or Ollama. Your provider bills you directly, with no markup from us, on every plan.
Open-source core
Agents run on OpenClaw, an open-source runtime. What your agent does is never locked inside a black box.

Bring your own model keySecurity and guardrailsThe AI agent platform in detail

VQuestions

Before you start

VIBegin

Step one takes a minute.

Hire an AI agentBrowse roles