Get your assistant off the ground.
Hands-on help getting Hermes running, understanding where it lives, and working through setup and authorization. We start with your equipment and needs—not a shopping list.
Hands-on AI & Hermes agent setup
AI, set up for
your actual life.
You have things you’d like help with. I help you turn that into a working AI assistant—with a setup you understand and access you’re comfortable giving.
Free 30-minute introductory conversation. New to AI agents? That’s a perfectly good place to start.
01 / From curious to using it
You don’t need to figure
it all out on your own.
Choosing tools is only part of it. The real work is getting them running, connecting the right things, and knowing what to do next. That’s where I can help.
Give it something useful to do.
A daily briefing. Research you can ask for. A way to turn actionable email into tasks. We choose a focused starting point and check how it works before adding more.
Understand what you’re handing over.
What can the agent access? What still needs your approval? What happens when a connection expires? I help you make those decisions rather than clicking through them for you.
02 / Jeremy’s experience
A friend who wanted less to manage—and someone to help him get started.
“He didn't just hand me a tool and leave; he made sure it was working and that I understood what I had.”
Shared with permission.
Jeremy wanted to reduce the mental load of managing multiple responsibilities. I introduced him to Hermes and worked alongside him to get the foundation running: a Raspberry Pi, Tailscale and the agent itself.
He now reports using it for daily briefings, financial tracking, calendar integration and research. What mattered wasn’t just getting software installed. It was taking the time to understand what he wanted from it.
More from Jeremy—and what we learned
“What I appreciated most was that Gary took the time to understand what I was actually trying to accomplish — not just technically, but in my life.”
Authentication took more effort than expected. Jeremy’s advice: handle authorization earlier and document the data structure before automating around it. He feels able to handle reauthorization and data corrections, while still wanting help with infrastructure issues.
These are Jeremy’s reported experiences, not measured performance results. The content pipeline was not live when he gave this feedback.
03 / How I approach it
One useful task.
A foundation
you can build on.
I’d rather help you get one thing working well than leave you with a collection of tools you don’t know how to use.
Find your starting pointUnderstand your situation.
What are you trying to make easier? What do you already use? We work out where an assistant could help—and where a simpler solution makes more sense.
Build the foundation together.
Set up the agent, work through the connections it needs, and agree on access boundaries. You should know what’s being connected and why.
Try a real task. Learn the basics.
Check the result together. Practice using the assistant and identify what you can handle yourself, what still needs review, and when to ask for help.
The person behind the setup
Hi, I’m Gary.
I like making
this stuff useful.
I work at the intersection of code, automation and governance, risk and compliance. I build with AI, use a personal assistant in my own workflows, and help other people get started.
That combination shapes how I work. I care whether the system does something useful, but also what it can access, what needs a human decision, and whether you understand it.
You don’t need to arrive with the technical answer. Bring the thing you’re trying to make easier.
04 / Work behind the offer
I build with it, too.
A few examples of the practical systems and product decisions behind my approach.
Personal AI / Everyday use
A personal assistant
with clear boundaries.
I use Hermes to turn actionable email into tasks and help with development and troubleshooting across my computers. I chose to separate its host account, keep administration private, and exclude credentials from configuration backups. It’s a working part of my setup—not just something I recommend.
The decisions behind the setup
A separate, non-admin host account. Private-network administration with a separate login. Allowlisted configuration backups that exclude credentials and private history.
I chose the goals and access boundaries; AI agents helped implement and troubleshoot them. Source notes document an email-to-task run and one verified remote connection. They do not establish availability across every machine or amount to a full security audit.
Workflow automation / GRC
From scattered reports
to a reviewable workflow.
In a large-scale access review program, I used Claude to help build a tool that consolidates reports, tracks outstanding removals and prepares follow-up notifications. I review the preview before triggering the send.
What gets automated—and what doesn’t
The tool combines campaign reports, shows pending and completed removals by application, and cross-checks entitlement inventory against campaign data. I remain responsible for validation and notification decisions.
It reduces manual effort in my workflow; no measured time-saving figure is claimed here. Employer names, internal systems and proprietary data are intentionally excluded.
Product development / In progress
Treat Dash.
A little play.
A lot of deliberate choices.
An offline Flutter/Flame mobile game: three backyard lanes, treats to collect and obstacles to avoid. I guide the product and play feedback; AI agents contribute implementation and tests.
Local progress. No account requirement. No tracking or ads in the free-game scope.
See more of the game
The full frame and crop come from development browser build 9efc080. This is an in-development project, not phone footage or a released app.
Before you start
A few fair questions.
What is an AI agent?
An AI agent is software that can use tools to help carry out tasks, rather than only replying to questions. What it can do depends on its setup, connected services and permissions. It still needs clear instructions and human review.
Hermes is the agent I helped Jeremy get started with. You don’t need to know the terminology before asking for help.
Do I need to be technical?
You don’t need to know how to set up an agent before asking for help. We start with what you want to do, explain the choices, and work through the technical parts together. You’ll still make decisions about your accounts and access.
Do I need a Raspberry Pi?
Jeremy’s setup used one; that doesn’t mean yours has to. We would first look at your existing equipment, the software requirements and whether you need something running all the time. Don’t buy hardware just to start a conversation.
Will it need access to my email or other accounts?
Only if the task calls for that integration and you decide to connect it. Start with the access the task needs. Connecting an account is a decision to understand, not a step to rush through.
Is this a set-it-and-forget-it system?
No. Connections can expire, services change, and AI can make mistakes. Part of getting started is understanding what needs review and how to recognize when something needs attention. Ongoing support, costs and scope would need to be agreed separately.
Your starting point
What’s one thing
you’d like
off your plate?
Start with a free 30-minute introductory conversation. We’ll talk about what you’re looking to do and how I can help. You don’t need a technical brief.
This is a conversation, not a hands-on setup session. If you’d like to take it further, we’ll agree on the next steps and any costs before starting further work.
- What are you trying to make easier?
- What have you tried so far, if anything?
- What would a useful first result look like?
Scheduling opens Cal.com. Email opens your email app. Please don’t include passwords or sensitive account details in your message.