CloudBot.ai

FAQ & Setup Guide

What the CloudBot harness is, how your agents plug into it, and the step-by-step setup guide.

The CloudBot harness — in plain English

What is a harness?

An AI agent is like an employee: it can talk and do work. A harness is everything around the employee — the phone system, the receptionist, the filing cabinet, the shift schedule. CloudBot is the harness for your AI agents: it gives them a voice that is always on, routes work between them, keeps every result from getting lost, and lets them reach you on their own when it matters.

Do I have to use your agent?

No. Every CloudBot harness ships with a default voice agent so it is useful from minute one — configure its personality, voice, and memory in the dashboard. When you are ready, swap it, run several side by side, or connect your own. The harness does not care whose face is on it.

What can I plug in as a backend agent?

Anything that can receive a message and send a reply: OpenClaw instances, n8n flows, MCP clients, Zapier or Make hooks, or a script on a laptop. Pairing is a code you paste; you approve the agent once and can revoke it anytime from the dashboard.

What happens when my backend agent is offline?

Tasks wait in the queue, and your dashboard shows each agent's last-seen time so you know who is alive. If you ask for an offline specialist, your voice agent tells you honestly instead of pretending. When the agent comes back, its finished work is still delivered — completed results are never dropped, even if CloudBot was unreachable when they finished.

Can my agents keep their own credentials — Google, email, everything else?

Yes — that is the design. Your backend agents do their own logins with third-party services; CloudBot never sees, brokers, or stores those credentials. And CloudBot never sees your agent's own login token either — it stores only a one-way hash, so it can verify the agent without ever knowing its secret.

What is a card?

A silent, glanceable status note your backend agents file: done, failed, or needs your decision. Cards never buzz your phone — the eye is the filter. Read them when you want, then turn one into a conversation with a tap, or have your agent call you about it.

Can my agent actually call me?

Yes. Ask your voice agent to call you tomorrow at six about the launch prep, and at that time your phone rings like a phone call — the agent answers already knowing why it is calling. If you miss it, your agent brings it up naturally the next time you talk.

Can I switch agents mid-conversation?

Yes. Ask for another agent by name and the harness hands off the call, carrying a short structured summary — the goal, decisions so far, open questions — so the new agent picks up right where the old one left off.

Does my voice agent read specialists' answers word-for-word?

No. The specialist sends its full written answer; your voice agent understands it and explains it naturally, in its own voice, in the context of your conversation. The frontend agent is the narrator — the specialists are the researchers.

Who owns my notes, tasks, and files?

You do. Your workspace content belongs to you. Backend agents can create and update things — and cancel their own tasks — but only you can delete anything.

For a visual tour of how the pieces fit together, see how multiple agents work together.

1) Create your CloudBot account

  1. Go to cloudbot.ai/signup and create your account.
  2. Confirm your email if prompted.
  3. Log in and open your Dashboard.

2) Copy your CloudBot API key

  1. In Dashboard, open Plugin or API Keys.
  2. Copy your tenant key (starts with cb_live_).
  3. Keep it private, this key connects your OpenClaw backend to your CloudBot tenant.

3) Install the OpenClaw CloudBot plugin

Run this on the machine where your OpenClaw instance is running.

openclaw plugins install @timekeeperbank/openclaw-cloudbot

4) Add CloudBot config to openclaw.json

Use your tenant API key from the dashboard.

{
  "extensions": {
    "cloudbot": {
      "gatewayUrl": "https://rtc.cloudbot.ai",
      "apiKey": "cb_live_YOUR_TENANT_KEY",
      "allowedAgentIds": ["main", "voice-agent"],
      "channelId": "cloudbot",
      "timeout": 130
    }
  }
}
  • gatewayUrl: CloudBot voice gateway.
  • apiKey: your tenant key from dashboard.
  • allowedAgentIds: optional, recommended for tighter routing control.

5) Configure your dashboard

Agent

Set agent name, greeting, system prompt, memory, and personality (Soul).

Voice

Choose TTS voice and optional start/interrupt sounds.

Model

Set your preferred primary and fallback model behavior.

Notifications & Settings

Set timezone, usage alerts, retention, and contact channels.

6) Verify connection

  1. Restart OpenClaw after plugin config changes.
  2. From Dashboard, check status in Overview.
  3. Run a test session and confirm voice + tool routing works.

Plugin & connection questions

Where do I find my API key?

In your CloudBot Dashboard under API Keys. It is tenant-scoped and usually starts with cb_live_.

What if plugin install fails?

Confirm OpenClaw is up to date and retry. Then verify the plugin name and network access from your server.

Do I need to restart OpenClaw?

Yes, restart after adding or updating plugin config so settings are loaded.

Can I restrict which agent receives CloudBot traffic?

Yes. Use allowedAgentIds in plugin config.