One-on-one agent setup calls & workshops
Get ready to put your agent to work.
Do the setup before we meet so we can spend our time on your business. Bring one goal, one approved example, and one agent you can use on your computer and phone.
Required before we meet
Finish these five steps.
-
Confirm your agent and paid account.
Use the agent named in your session instructions, or choose one with your facilitator: Muse, dot in ChatGPT, or Claude Duet. You only need one track.
For Dennis’s hands-on setup sessions, arrive with a paid account for your chosen agent, ready to use. That is our preparation expectation; vendors may also offer free access. Before subscribing, confirm that your actual account, region, device and invitation qualify for the feature we will use. A paid plan alone does not unlock every agent. Don’t buy all three.
Ready when: you can open your chosen agent and get a reply. If you see a waitlist, missing feature or administrator request, report the exact blocker before the call.
-
Set up your computer and phone.
Have a laptop or desktop for the setup work, plus your phone. Update the apps in your chosen track and sign in to the same agent account on both devices. Check that you can access your verification email and your normal two-step sign-in method. Keep those codes private.
Ready when: you finish the phone-to-computer test below. If your device is unsupported or managed by your employer, or you need an accessibility accommodation, tell the facilitator ahead of time so we can agree on a supported setup.
-
Get your GitHub account ready.
GitHub holds the instructions and files we may use together. Create a GitHub Free account and verify your email, or sign in to the account you already use. GitHub Copilot and a paid GitHub plan are not required for this preparation.
Open the repository link supplied for your session and read its README, the starting instructions. If it is private, accept your invitation and confirm the files open. If no link was supplied, include that in your readiness reply. Use your organization’s normal approval process for work repositories.
Ready when: you can name your GitHub username, open the intended repository while signed in, and find its README. Don’t send a password, access token, or verification code.
-
Make Wispr Flow type what you say.
Install and set up Wispr Flow on the device you will use to explain your work. Free access is enough for this short test. Check the supported devices, allow the microphone and required text-entry permissions, and choose your shortcut or mobile keyboard. Review the privacy choices during setup.
In a blank note, dictate: “My business helps [audience]. Before [date], I want to finish [one useful task]. My agent’s name is [name].” Read the text and correct names, numbers and dates before you paste or send it.
Ready when: your words appear as editable text in a note. If Flow is blocked, report the device and failed step; type your brief or use approved device dictation for the exercise while setup is resolved.
-
Bring one real business task.
Write one measurable goal, the audience it serves and what a useful first result looks like. Bring one asset you may use: a public business page, approved recording, review or work sample, with its source link. For example: turn one approved interview into a draft FAQ for one audience.
Keep the first task small enough to inspect together. A useful starting request is: “Using only this approved source [link], draft five questions our customers ask and short answers. Link the source, flag unknowns, and leave the draft here for me to review.”
Ready when: you can open the source and explain how you will judge the draft. Bring a public or made-up sample as a backup if your real material cannot be shared.
Choose one track
Use the same agent on both devices.
Start from these official links so you install the right product. Confirm current availability before you pay or install.
Track A
Muse Meta’s personal agent
Start at muse.ai. Install one agent app on your phone: Muse. Use its official mobile and Mac setup links with the same account. This is Meta’s personal agent.
If you use Windows or another desktop platform, have your facilitator confirm a supported computer route before choosing this track. Do not assume a Mac app will run on your computer.
Test: open your existing Muse on your phone and computer, then complete the continuation test below.
Track B
dot Your agent in ChatGPT
Follow OpenAI’s dot setup and availability guide. Install one agent app on your phone: ChatGPT, plus the ChatGPT desktop app on your computer. Sign in to the same account and open your dot on each device.
Check that dots are enabled for your actual account. Rollout and workspace settings apply; paying alone does not guarantee access. The guide also supports a desktop browser. Mobile web does not support dot setup.
Test: continue with the same dot on your phone and desktop. A regular ChatGPT chat by itself does not pass this track.
Track C
Claude Duet Invitation and eligibility required
Check the official Duet guide and your invitation. Confirm that Duet opens in the account tied to that invitation; a normal Claude chat is a different experience.
For our hands-on setup as of 3 November 2026: install both Claude and Claude Duet on your phone, plus Claude Desktop on your computer. Use Claude’s official downloads and the Duet guide’s official app links. Sign in with the account tied to your Duet invitation.
This is our session’s teaching setup. The ordinary Claude phone app is not a universal prerequisite for Duet access. Duet also has an official browser option; use that if your facilitator agrees on a browser-based session.
Test: Claude and Claude Duet open on your phone, Claude Desktop opens on your computer, and you can continue the same Duet task on phone and computer using its supported app or browser route.
The phone-to-computer test
- On your phone, ask your chosen agent: “Make a three-item checklist for our agent setup call. Use no private data and take no action outside this conversation.”
- Open that same agent and task on your computer. Confirm you can see the checklist.
- On the computer, ask: “Add ‘bring my charger’ to that checklist.” Return to your phone and confirm the update is there.
Pass: you can see the same task and its update on both devices. If the apps open different accounts or the task does not carry over, report that step. Installed icons alone are not enough.
Setup links reviewed 3 November 2026. Products and access change; use the linked official instructions and your facilitator’s current session note.
Have these within reach.
- Your charged computer and phone, chargers, working internet, and headphones if you are in a shared space.
- Access to your own sign-in and two-step verification, your GitHub username and repository link, and your tested agent account.
- Your goal, approved source link, first task, and a saved public or made-up fallback sample.
Keep credentials, private messages and sensitive customer records off shared screens. Connect Drive, a CRM, project tools or an inbox only if the chosen task needs it and the account owner approves the scope. Private inbox access, shared memory and schedules are optional later steps.
If access, Wi-Fi or installation fails, use your saved sample or pair with a prepared participant. Keep the blocked setup step visible and complete the work that is still possible.
Keep the workshop resources handy.
Workbook library · Participant workbook · Facilitator guide · Current skill setup guide · Workshop formats and examples
For strategy, use Goals, Content, Targeting: one audience, one useful offer and a measured next action.
Before the call
Reply READY or BLOCKED.
Send this in your existing booking or team thread to your named facilitator. Include the relevant setup teammate. Use the contact supplied with your session instructions.
READY / BLOCKED — [session date] Computer + OS: [device / version] Phone + OS: [device / version] Agent + paid plan: [Muse / dot / Duet; access working?] Apps + continuation test: [pass / exact failed step] Duet only: [Claude + Claude Duet on phone; Claude Desktop ready] GitHub: [username; session repository opens / blocker] Wispr Flow: [dictation test passed / blocker] Business goal + first task: [one sentence] Approved source: [public link or description; no private records] Needed work tools: [Drive / CRM / project tool / none yet] Blocker: [step; expected result; what happened]
A redacted screenshot can help with a blocker. Leave out passwords, tokens, sign-in codes and private invitation links. A blocked feature is useful to report before the call; don’t mark it READY just because the app is installed.
Leave with one result you checked.
Save the draft, its sources, your corrections and the next action. If you install a skill, start a fresh task through the product’s supported controls and inspect its output. If you connect a work account, test only the approved input and output. Treat a recurring workflow as active only after you have seen a run and checked the result.
Keep a record of what you built.
Use this table to record the stage you reached and the next step.
| State | Evidence to keep |
|---|---|
| Public exercise | A sourced draft with the goal, audience and unknowns clear. |
| Installed skill | Supported install completed, actual selected instruction identified, new task output inspected. |
| Connected workflow | Scoped account permission plus a safe test with its approved input and output. |
| Recurring operation | Supported trigger, approved scope, an observed execution and a reviewed result. |
For organizers and facilitators
Before sending this checklist, confirm the agent, exact apps, supported devices, required plan/access, repository and first exercise. Our current phone setup is Muse alone, ChatGPT alone, or both Claude and Claude Duet for the Claude track. Recheck the dated app list and desktop route before each session. Rehearse the full path on desktop and mobile, including workbook downloads. Keep one resource-page owner and a current workbook edition. Event dates, venues and participation requirements belong in the session instructions.
This checklist is shared across workshop groups, including Hypergrowth. For a skill exercise, record the actual instruction/source used, what happened, what failed and what you changed. Keep the completed installation steps and inspected task output as evidence; a model’s “skills loaded” response alone does not establish installation.
Afterward, record what people completed and where they got stuck. Use Metrics → Analysis → Action to fix the source instructions and dependent materials before the next session.