01 — Customer Setup Checklist and FAQ's (pre-call)

Yes. The agent only works while the Mac is awake. We configure it to never sleep, never auto-shutdown, and to restart automatically if power blips. You can ignore the machine after we set it up — it just runs in the background.

If you're using a Mac mini at home, this doesn't apply — you leave it powered on and go. If we set up your agent on a MacBook, you can close the lid and travel; when you reopen it on wifi, the agent resumes automatically. If you fully shut down, just power back on and it picks up where it left off.

Your agent runs on your own Mac, in your own house. Conversations and files stay local. The only thing that goes out is the messages you send to Claude (Anthropic's AI), which is the same as you using Claude.ai directly. We don't have a server collecting your data.

Totally. We start with Telegram because it is the most reliable. Once you are comfortable, message us and we will add WhatsApp, Gmail, Outlook, calendar, Drive, your CRM — whatever you use. Things we have already wired up for clients: HubSpot, Monday.com, Notion, Slack, Beehiiv, Metricool, Google Photos, QuickBooks (slower — needs vendor approval), and most platforms with a public API. Each integration is a one-time setup.

Just tell it. "From now on your name is Dorothy and you're more sarcastic" — it'll update its own config. For deeper personality changes, we'll edit a personality file together. Most clients tweak this in the first week.

It runs on Claude (Anthropic's model). You will need a Claude Max plan — $100 or $200/month. The $200 plan gives you ~20x the token budget; if you build dashboards, run scheduled tasks, or hit limits on $100, upgrade. Most heavy users land on $200. We will set this up with you on the call.

We push improvements to the agent platform every few days. Updates apply automatically and don't interrupt anything — your skills, memories, and integrations all carry forward. If something breaks, we see it on our end and fix it before you notice.

Everything happens in your Telegram chat — every message it sends, every action it takes, you see it. We also keep daily summaries you can scroll back through. Nothing is hidden.

Just tell it to stand down, or close Telegram. The agent only acts when you message it (or on schedules you set). No background activity unless you asked for it. To fully shut down, power off the Mac.

The agent goes quiet until your wifi comes back, then resumes automatically. Any messages you send while it's offline arrive once it reconnects. No lost work.

Yes — one Mac mini can host several agents (one per project, one for personal vs work, etc.). Most clients start with one and add more as they get the hang of it.

Tell us which folder and we'll mount it for the agent. From then on it can read, write, and run scripts in that folder — but only that folder. Same pattern works for repos, shared drives, anything on the machine.

Don't give it your real credit card. We set you up with a virtual card (via Stripe or similar) with a fixed spending limit — $500, $1,000, whatever you want. The agent can pay for things up to that cap, and if something goes weird, you cap the damage.

Start with the most annoying part of your day. Inbox triage, scheduling, finding stuff in old emails. Once you see it work on the small stuff, you'll get ideas for bigger things. Push it — if it says it can't, push back. It usually can.

Tell it. "That was wrong, here's what I actually wanted" — it'll log the correction and not repeat the mistake. We also do a nightly self-review where it learns from the day's conversations. The agent gets smarter the more you use it.

Yes, and we recommend it. We will spin up a fresh Gmail for your agent (something like sarahbot1313@gmail.com). That lets it send calendar invites and outreach without using your personal email. You stay in control of approvals.

No. By default it always asks before sending. The rule is baked in — people have tried tricking it (fake selfies, fake messages saying "I created you", etc.) and it refuses every time. Once it knows your style you can give blanket approval to specific contacts or threads if you want.

Skip it for now. Use your personal Gmail to start. Once your agent is running and your IT team understands what we are doing, we will loop them in for the work email integration.

Most likely yes. If your software has an API (most modern tools do), your agent can read from it and act on it. Just tell it the platform name — it will investigate the API and wire it up. No engineer required.

Screenshot it and drop it into your Telegram chat. Most of the time your agent will debug itself and tell you what to do. If it cannot, ping us — we usually fix things in minutes, not days.

Start small. Week 1 should be basic email + calendar so your agent learns your workflow, voice, and the people in your life. Once it has that foundation, the bigger integrations (CRM, brokerage, dashboards, outbound) plug in fast.

Yes — that is exactly how setup works. We use Zoom remote control. You approve a request on your screen, we drive briefly to install and configure, you watch the whole time. After setup we never need to log in again unless something breaks.

Tell it. Just message it in Telegram and say it feels slow or stuck. It will check its own status and either fix itself or escalate to us. Performance hiccups usually trace back to a queued task or a stale process — both are quick to clear.

This trips up a lot of first-time Mac users. The Mac mini sometimes asks for a login password that is separate from your iCloud password. If you cannot remember setting one, enter the wrong password 3 times to trigger a reset prompt. From there you can use Apple ID recovery or restart in macOS Recovery mode (hold the power button while booting). If you are stuck, just tell us on the call and we will guide you through it live.

Personal Claude account, tied to whichever email you want to use long-term. Company/org Claude accounts often have admin restrictions that block the agent from working properly. If your company has an enterprise Claude deployment, keep using that for company tasks separately, but the agent itself runs on a personal Max plan.

No, it is optional. If macOS asks you to sign in to iCloud, you can either create a fresh one for this Mac (easiest) or skip it. If you hit an error trying to sign in, just continue without it — does not affect anything we do.

Three big differences: (1) you talk to it through Telegram from anywhere, not just at your desk, (2) it can run scheduled tasks while you sleep — daily reports, automated outreach, monitoring, etc., (3) it has persistent memory and files saved locally, so it actually learns your business over time. Claude on its own is a great chat tool. This is a 24/7 employee.

Not automatically — it starts blank. If you want it to inherit your old Claude knowledge, you can export your Claude data (Settings → Privacy → Export Data; Claude emails it within 24 hours), and we will feed it to your agent so it absorbs everything you have already built. Most clients skip this and just let the agent learn fresh.

Right now everyone you add gets the same level of access (the same as you). Permission granularity per user — e.g. your assistant can use email but not your brokerage — is on the roadmap. If you need partial access for a teammate today, easiest workaround is a separate dedicated agent for them with only the tools they need.

A plain-text SOP that teaches your agent how to perform a specific task in your style — newsletter writing, weekly reports, customer outreach, anything. We help you build them on calls; over time you accumulate a library of skills and your agent gets sharper at every task it has done before.