// CYNDRA_ONBOARDING / 02_DAY_OF_CALL

Day of the call

Your 90-minute onboarding call. By the end you'll have a working AI agent you can chat with from your phone. Here's exactly what we'll do, what you need ready, and how to get the most out of the time.

Have these ready before we start

15 minutes before the call, do this final check. If anything is missing, ping us and we'll move the call rather than burn the first 30 minutes troubleshooting.

On your Mac mini

  • Powered on, plugged in, connected to wifi or ethernet.
  • macOS updated to the latest version (System Settings → General → Software Update).
  • Zoom installed, opened once, granted Accessibility + Screen Recording permissions.
  • Telegram desktop app installed and paired with your phone.

On you

  • Your Mac user password written down (the one you set during onboarding — not your Apple ID).
  • A credit card for the AI subscription (Claude or OpenAI — we'll decide which on the call).
  • Your phone unlocked and within reach. Telegram open.
  • A quiet 90 minutes. Closed laptop, kids/pets handled, coffee made.
  • An idea of the most annoying part of your week. We'll wire your agent up to handle it on day one.

Optional but nice

  • A second screen or laptop. Lets you watch what we're doing on the Mac mini in Zoom while taking notes on your laptop.
  • Headphones. Better audio than your laptop speakers.
  • Your team's contact info, if you want anyone else added to the agent's group.
If something breaks pre-call, screenshot it and send it in chat. Don't try to fix it yourself — we've seen every error. Most of them resolve in 30 seconds once we see them.

The 90-minute agenda

Here's how the call actually flows. Times are typical, not strict — we'll adapt to what your business needs.

0:00 – 0:10

Hello + Zoom remote control

Quick intros. We confirm what your agent is going to focus on (the "most annoying thing" you came in with). You approve the Zoom remote-control request so we can drive your Mac. We do a 60-second sanity check that wifi, Telegram, and Zoom permissions are all good.

0:10 – 0:25

Install the Cyndra agent

We install the agent runtime on your Mac. This is the part where we're typing on your machine for ~15 minutes. You watch the whole thing — we narrate what each step does so you understand what's happening.

No action from you here unless macOS prompts for your password. When it does, you type it in.

0:25 – 0:40

Pick the AI plan + plug it in

We talk through your use case for ~5 minutes and pick Claude or OpenAI. Heuristics: Claude wins for writing, reasoning, and longer-form work; OpenAI wins for tool use, voice, and image generation. Most clients land on one or the other — a few want both.

You subscribe to the chosen plan ($100 or $200/mo). We connect your account to the agent and confirm the token budget is healthy.

0:40 – 0:55

Pair Telegram + first conversation

We create a Telegram group: you, your agent, and us (we'll leave once everything is stable, usually within a week). You say "hi" to your agent, it introduces itself, you teach it your name and what you do.

Then you give it the small first task — "remember that I prefer concise replies", "here's my title and company", "my work hours are X to Y in Z timezone". The agent saves these and uses them forever.

0:55 – 1:20

First real integration

We hook up the first real tool based on what you came in for. Most common starting point is Gmail or Google Calendar. We walk through the OAuth approval, then you ask the agent to actually do something — "what's on my calendar today?", "summarize my unread emails", "draft a reply to Sarah but don't send yet".

You watch it work. You correct it when it gets something wrong. The agent learns from each correction.

1:20 – 1:30

Wrap-up + what's next

Quick recap: where the agent lives, how to reach it, how to reach us, how the next week works. We point you at the first-week guide so you know what to try next.

We schedule a 30-minute follow-up for ~7 days out to add more integrations once the agent has learned your basics. Done.

What we won't do on the first call

Day-one is foundation. We deliberately keep it tight so the agent works immediately and you can start talking to it tonight. Bigger stuff comes in the second call once the basics are nailed.

  • Connect your CRM (HubSpot, Salesforce, Monday, etc). Comes in week 2 once we know your sales workflow.
  • Build custom skills or SOPs. We capture them as you describe what you want — they get built over the first 2 weeks.
  • Set up scheduled tasks (daily reports, weekly summaries, automated outreach). Easier to add once we know your rhythm.
  • Make the agent fully autonomous. By default it asks permission before sending emails, paying for things, or making changes. We tune that down only after you trust it.
  • Add team members. We can if you want, but most clients add their assistant or partner in week 2 after they've personally tested it.

How to get the most out of the 90 minutes

  • Be specific about what's annoying you. "Email is a mess" → we can work with that. "I want it to handle emails like Sarah does" → even better.
  • Ask dumb questions. There are no dumb ones. The customers who get the most out of week 1 are the ones who interrupt to ask why we're doing something.
  • Don't take notes — we'll send a recap. Just watch and absorb.
  • If you're confused about something the agent did, tell it on the spot. "Why did you do X?" — it will explain. That's how it learns your preferences.
  • Bring real work. The agent learns fastest when you give it actual tasks, not test prompts.

Common snags during install (and how we handle them)

These come up often enough that we want you to know they're normal — not signs of a broken install.

  • macOS asks for your password 4-6 times during install. Normal — each ask is for a different system permission. Just type it.
  • Browser prompts you to allow OAuth for Gmail/Calendar/etc. Click Allow. We'll narrate each one.
  • Telegram group invitation gets buried under other notifications. Check Telegram settings → Notifications and bump the invite-priority.
  • First message to the agent takes 10-30 seconds to respond — the model is warming up. After that it's near-instant.
  • If your wifi blips during install we just resume from where we were. No data loss.

After we hang up

Once the call ends, the agent is yours. Here's what to do in the first hour after.

  • Open Telegram on your phone. Send your agent a message — "hi", "what can you do?", anything.
  • Try one real task you'd normally do yourself. Email triage. "What's on my calendar tomorrow." "Find that thread from Maria last month."
  • When it does something well, tell it: "that was perfect, do it like that always". When it gets something wrong, tell it that too.
  • Don't worry about breaking it. You can't. Worst case it does the wrong thing once, you correct it, it adjusts.
  • Keep the Mac mini powered on. Don't shut it down. The agent only works while the Mac is awake.
Day one homework: push the agent. Try to make it do something hard. Try to break it. The faster you stress-test it, the faster it adapts to you. Plus you'll spot the limits and we can address them in the follow-up call.

// NEXT

Next: your first week with the agent

What to try in the days after the call, how to teach your agent your style, and what to ping us about vs. what to handle yourself.

03 — First week with your agent