💾 Welcome.exe - Michael Costea Portfolio

HEAD OF TECH, AI & SYSTEMSAI, SYSTEMS & OPS

LOADING BUSINESS_OPERATING_LAYER... [OK]

I build AI-powered business operating systems that turn messy workflows into visible, scalable, human-reviewed systems.Visible, reviewed AI systems.

I am Michael Costea, currently leading technology, automation, AI, digital infrastructure, and growth systems across a multi-brand electrical and energy services group.

My current mission is building a practical AI and workflow layer on top of core business systems so the company can grow faster, operate more efficiently, and deliver a better customer experience without unnecessary support-function headcount growth.

The goal is not AI for AI’s sake. It is a connected, visible business where sales, ops, finance, admin, and field teams work through clearer system architecture, review gates, and measurable operating loops.

Faster lead responseLess duplicate adminVisible handoffsHuman-reviewed automation

CURRENT ROLEHead of Tech, AI & Systems · All Electric Homes

Ready. System status: Scaling.AI: LIVEOPS: ONLINE

Programs

Pixel art coming soon graphic for more tools, guides, and experiments

Applications hidden until release.
Public webapps, games, demos, and internal tools are hidden until they are rebuilt to the MICHAEL OS 89 style guide and ready for public release.

My Computer - Current Operating Profile

Current Operating Profile

Pixel art My Computer graphic introducing Michael's current focus across AI, systems, strategy, growth, and automation

Name: Michael Costea

Current role: Head of Tech, AI & Systems — All Electric Homes

Location: Melbourne, Victoria, Australia

Direction: CTO-track systems and technology leadership across multi-brand electrical and energy services.

What I am building now

A scalable AI and workflow layer on top of core business systems. The layer is designed to connect marketing, lead generation, sales, customer journey, operations, finance, admin, internal tooling, and job completion into a clearer system architecture.

More business context visible across teams and systems.
Less duplicate entry, fragmented handoffs, and manual admin.
Faster lead response, quote preparation, and conversion flow.
More consistent customer experience from first touchpoint to job completion.
Automation that is reviewable, measurable, and safe by design.

Resume.doc - Michael Costea

Michael Costea

Head of Tech, AI & Systems · CTO-track Technology Leader · Business Systems & Automation Operator

Melbourne, Victoria, Australia · costea.michael@gmail.com · LinkedIn

Pixel art resume graphic for experience, skills, leadership, and impact

Executive Profile

Technology and operations leader currently owning the design and execution of technology, automation, AI, and growth systems across a multi-brand electrical and energy services group. My career has moved from electrical delivery, sales, operations, process improvement, and marketing into full ownership of systems, automation, digital infrastructure, and technical strategy.

I am building a scalable AI and workflow layer over core business systems so the company can grow faster, operate more efficiently, and deliver a better customer experience without unnecessary headcount growth in support functions. This is a CTO-track operating role: clear system boundaries, practical automation, measurable workflows, and human review where risk, quality, or customer trust matters.

Current Strategic Focus

Own and manage multi-brand websites, digital infrastructure, lead systems, and internal tooling.
Design and optimise lead generation, conversion flow, lead routing, and qualification systems.
Build and deploy AI-powered automation using agent workflows, including OpenClaw and related local systems.
Integrate marketing, sales, customer service, operations, finance, and admin workflows into a clearer architecture.
Oversee customer journey from first touchpoint to job completion.
Reduce manual admin workload and enable scalable growth without increasing operational overhead.

Experience

Head of Tech, AI & Systems — All Electric Homes

Feb 2026 - Present · Melbourne · On-site

Pixel mini Michael stabilising production systems for AEH

Lead the design and execution of technology, automation, and growth systems across a multi-brand electrical and energy services group. Own digital infrastructure, websites, lead systems, AI-powered automation, customer journey systems, internal tooling, process automation, and technical strategy.

Impact: reduced manual admin workload, improved lead response speed and conversion flow, centralised systems across multiple brands, and enabled scalable growth without increasing operational overhead.

Business Development & Marketing Manager — Want A Heat Pump / Want A Sparky / All Electric Homes

Feb 2024 - Apr 2026 · Australia

Pixel mini Michael driving growth systems

Drove B2B and partner client relations, revenue generation, client acquisition and retention, digital strategies, website management, user-metric analysis, lead generation pathways, IVR overhaul, online review automation, chatbot integration, targeted email campaigns, partner programs, and community master class education.

Sales Manager — Want A Heat Pump

Nov 2023 - Apr 2026 · Melbourne

Pixel mini Michael in sales and customer journey mode

Worked with authorised retailers and delivered heat pump upgrades across residential and commercial sectors under the Victorian Energy Upgrade Program. Managed CRM-driven sales activity, conversion flow, customer relationships, and strategic marketing support.

Electrician — Want A Sparky

Apr 2023 - Jun 2024 · Mornington / South East Melbourne

Delivered domestic, commercial, and emergency electrical services across Want a Sparky, Want A Charger, Want a Heat Pump, and All Electric Homes. Supported the business transition toward electrification and greener energy services.

Electrician — Wired By MJD

Jun 2022 - Apr 2023 · Melbourne

Pixel mini Michael with smart-home systems

Worked across smart home and integration electrical projects including custom builds, undergrounds, fault finding, renovations, high-end A/V and cinema installs, large rack builds, data, security, TV, KNX wiring, C-Bus diagnosis, custom lighting, and architectural builds exceeding $4M.

Electrician — Ezzy R Electrics

Feb 2018 - Jun 2022 · Melbourne

Pixel mini Michael troubleshooting electrical faults

Built broad electrical trade experience across intercoms, telecommunications, A/V installation, fault finding, medical installations, renovations, custom installs, residential/commercial maintenance, network configuration, pool, garden and deck lighting.

Operations Manager / Owner — Bloom Coffee Bar

Jul 2017 - Feb 2018 · Carlton

Pixel mini Michael running a cafe counter

Assisted in establishing and operating an owner/operator cafe, developing hands-on experience in people management, operations, customer experience, and small-business execution.

Optus — Sales Operations Support Manager / Process & Content Manager / Business Process Specialist / Retention Consultant

Aug 2010 - Jun 2017 · Melbourne

Pixel mini Michael handling large-scale support processes

Managed retail support operations across 352 stores, maintained knowledge bases serving 5,000+ employees and 500,000 monthly visits, led process redesigns, support operations, credit policy improvements, training, communications, and large-scale change management.

Selected impact: 74% increase in work outputs over 12 months, removal/rewrite of hundreds of duplicate processes, support-ticket reduction, improved knowledge search accuracy, and redesigned operational reporting.

Core Capabilities — AI, Systems & CTO-Track Execution

AI Strategy & Roadmapping

identify high-leverage automation opportunities, prioritise by risk/ROI, define AI adoption roadmap, and translate business problems into practical technical systems.

Agentic Workflow Automation

OpenClaw/Hermes-style agents, scheduled automations, watchdogs, multi-step task execution, human-in-the-loop review, fail-closed workflows, and escalation paths.

Business Systems Architecture

connect marketing, sales, CRM, ops, finance, admin, field delivery, reporting, and customer journey into clear system boundaries with orchestration between them.

Digital Infrastructure Ownership

multi-brand websites, domains, hosting, analytics, conversion pathways, landing pages, lead capture, webapp prototypes, and technical vendor/tool rationalisation.

Lead Gen & Conversion Systems

lead routing, qualification, response-speed improvement, nurture/follow-up automation, review management, IVR/chatbot improvements, and conversion-flow optimisation.

Data, Reporting & Visibility

operational dashboards, exception surfacing, KPI/report redesign, source-of-truth thinking, clean data capture, workflow measurement, and decision-support views.

Process Automation & Admin Reduction

remove duplicate entry, simplify fragmented handoffs, automate routine admin, standardise operating procedures, and reduce support workload without hiding risk.

Integration Methods

API-first integration, browser automation as fallback, desktop/manual last resort, structured handoff design, audit trails, and measurable control points.

Customer Journey Systems

first-touch to job-completion visibility, escalation handling, customer communications, partner/client workflows, technician support, and consistent service experience.

Change & Knowledge Systems

training packs, internal communications, knowledge-base design, process rollout, adoption support, stakeholder alignment, and team enablement.

Operations & Field Context

licensed electrical background, heat-pump/electrification context, smart-home integration, customer-facing delivery, field constraints, and practical service-business execution.

Leadership & Commercial Judgment

bridge sales, operations, technology, marketing, partners, and leadership; balance automation with quality, customer trust, risk, approvals, and growth strategy.

What I Build - AI Business Operating Systems

What I Build

Pixel art workflow systems graphic showing CRM, automation, dashboards, operations, validation, and visibility

I build practical AI-powered operating layers for growing service businesses: lead flow, customer journey, operations, reporting, knowledge, and admin work made visible and reviewable.

Lead Response & Customer Journey

Lead intake, qualification, quote prep, follow-up, review requests, warranty loops, and handoff summaries so fewer opportunities go stale.

AI Agent Workflows

OpenClaw/Hermes-style agents with bounded tools, Telegram accountability, logs, receipts, review gates, and fail-closed permission rules.

Dashboards & Exception Radar

Operational views that surface overdue jobs, stale leads, missing payments, review gaps, bottlenecks, and evidence links — not vanity charts.

CRM / Ops / Admin Integration

Clear workflow boundaries between sales, marketing, admin, finance, field teams, reporting, and customer communications.

Knowledge & Process Systems

SOPs, repeated-question capture, training packs, internal answers, source-of-truth cleanup, and process simplification before automation.

Multi-Brand Digital Infrastructure

Websites, landing paths, lead capture, analytics, content systems, release gates, and technical vendor/tool rationalisation across brands.

Operating Principles

Clarity before automationAPIs firstHumans keep judgmentEvery workflow needs receiptsSmall proven loops first

Projects - AI Systems Portfolio

Project Portfolio

Pixel art projects graphic for builds, experiments, systems, and impact

Positioning: this is a case-study style portfolio, not a random app list. Each bucket is framed around the business problem, the system being built, and the operating result it is designed to create.

Public demos remain hidden until release.
The showcase explains the systems being built or led behind the scenes without exposing internal tools, customer data, or fragile prototypes. Scroll inside this window to see the case-study buckets.

AI-firstNo duplicate agent bucketsBusiness outcome ledHuman-reviewed automation

Case-study lens: messy workflow → designed operating loop → visible owner/status/evidence → human review where trust, money, quality, or customers are involved.

Latest Build Notes — updated from project memory

Remote Autonomy Reliability Pass

Cleaned up noisy health warnings in the agent operations layer: long Telegram reports now split safely, memory/task bridge calls accept the current automation format, and stale repeated error logs were archived instead of polluting live health checks.

agent ops · health checks · Telegram updates · observability

AI Help Guide 3: First Safe Workflow

Added a beginner-friendly tutorial for turning one private repeatable task into a safe AI workflow: write the goal, test with harmless data, add approval gates, save the playbook, and only then reuse it.

AI education · beginner workflow · human approval gates

Operational Backup & Memory Receipts

Continued the pattern of backup snapshots, memory notes, and verifiable receipts so builds can be reviewed later without relying on vague “it should be fine” status updates.

backups · memory spine · audit trail · build receipts

1. AEH AI Operating Layer

Strategic AI and workflow layer across sales, operations, finance, admin, reporting, customer experience, lead flow, and job completion. The goal is a more visible business where repeatable work is automated, exceptions surface early, and humans stay on judgment.

Maps core workflows and handoffs.
Defines automation boundaries, approval gates, and escalation rules.
Turns fragmented work into measurable operating loops.

AI strategy · business systems · operating leverage

2. Customer Journey & Revenue Automation

AI-assisted customer and lead workflows: fast response, qualification, quote preparation, follow-up, stale-lead detection, review loops, and handoff summaries for staff.

Lead intake → CRM context → next best action.
Quote/admin prep with missing-info flags.
Human approval before customer-impacting sends.

CRM · lead conversion · customer experience

3. Agent Infrastructure & Shared Context

Personal and business agent stack using OpenClaw, Hermes, Telegram, dashboards, scheduled jobs, watchdogs, memory, and shared context across phone, Windows control plane, and execution machines.

Visible task receipts instead of silent automation.
Shared memory/state so agents do not restart from zero.
Phone-first command path with reviewable outputs.

OpenClaw · Hermes · Telegram · memory · watchdogs

4. Data Visibility, Reporting & Exception Radar

Dashboard and reporting patterns that turn scattered operational data into clear action lists: overdue work, missing payments, stale leads, customer risks, review gaps, and process bottlenecks.

Exception-first reporting, not vanity dashboards.
Evidence links and audit trails for decisions.
Cleaner data capture at workflow source.

dashboards · KPI redesign · exception surfacing

5. Multi-Brand Digital Growth Systems

Digital infrastructure across Want A Sparky, Want A Heat Pump, All Electric Homes, and personal brand surfaces — websites, lead pathways, content systems, analytics, conversion flows, and safe public-release controls.

Brand-specific journeys with shared operating logic.
Lead capture and conversion path ownership.
Public demos gated until stable and aligned.

web platforms · lead gen · analytics · GitHub Pages

6. Knowledge Systems & Process Modernisation

Process and knowledge-system foundation from Optus-scale work through current AI enablement: SOPs, support flows, training packs, internal answers, duplicate-process removal, and change rollout.

Knowledge bases designed for staff adoption.
Repeated questions converted into reusable SOPs.
Process simplification before automation scale-up.

knowledge ops · SOPs · change management · scale

AI Agents - Business Operating Leverage

AI Agents & Workflow Automation

Pixel art AI Agents graphic for tools, workflows, and automation

Ethos: build AI agents the way you would build a serious operating system for your life or business: visible, measured, permissioned, reviewable, and useful every day. The win is not replacing people with a chatbot. The win is removing repetitive drag, keeping context alive, catching exceptions early, and giving humans more leverage.

Most stock agent installs are impressive demos but weak operating layers. They can answer, code, and use tools, but they often lack durable memory, shared visibility, scheduled follow-up, Telegram accountability, watchdogs, dashboards, and a review loop. Our current stack is designed to close that gap.

OpenClaw

Local automation/control layer: terminal UI, gateways, Telegram channels, browser control, plugins, memory, sessions, and dashboard access.

Hermes

Planning and execution layer: coding, repo edits, scheduled jobs, local APIs, research, verification, memory retrieval, and Telegram updates.

Master Control

Visibility layer: projects, live services, scheduled tasks, all sessions, memory, Obsidian notes, and agent engagement from one cockpit.

Human Review Layer

Trust layer: approval gates, evidence, audit trails, fail-closed workflows, exception surfacing, and handoff back to a person.

Diagram of Michael Costea's current daily AI agent setup showing phone as remote entry point, Windows PC as central control hub, local OpenClaw Hermes and Paperclip agents, shared operating context, and a Work Mac Mini as secondary execution environment. — **Current daily setup:** phone → Windows PC control plane → local agents + shared operating context → Work Mac Mini execution environment. Click the diagram to open it full-size.

Concerns and pitfalls to respect

Silent failure: agents can look busy while missing the real business outcome. Track inputs, outputs, receipts, and verification.
Tool overreach: never let an agent delete, spend, message customers, or change systems without explicit permission and rollback paths.
Context rot: long chats drift. Use memory deliberately, compact sessions, and keep source-of-truth docs clean.
Model cost and rate limits: long tool loops burn tokens. Monitor usage and pick the cheapest model that can reason well enough for the job.
Browser automation fragility: APIs first, browser automation second, desktop/manual fallback last.
Trust gap: if a human cannot see what happened, why it happened, and what changed, the workflow is not production-ready.

Why this can become a cheat code

When agents are tracked and monitored properly, they become a compounding leverage system: they remember decisions, chase loose ends, prepare work, run checks, produce evidence, and keep pushing while you sleep or focus elsewhere. That is the real cheat code — not one magic prompt, but a monitored agent stack that keeps life and business workflows moving.

Build principles

Start with one painful repeatable workflow, not a vague “automate everything” goal.
Give the agent bounded tools, test data, and a clear success definition.
Make every workflow observable: logs, Telegram updates, dashboard state, and final receipts.
Keep humans on judgment, customer trust, finance, safety, legal, quality, and edge cases.
Scale only after the workflow proves it saves time without creating hidden risk.

AI Help - Start Here

AI Help: Everyday Starter Guides

Pixel art AI Help graphic for beginner-friendly guides and tutorials

This is the starting point for normal people who want a safe, simple first agent. No hype. No jargon wall. Learn what agents do, then follow the install path for Hermes or OpenClaw.

Pixel mini Michael high-fiving a robot for the AI Help starter path Start simple: pick one guide, test in private, and keep humans approving anything risky.

Learn
what agents do

→

Choose
Hermes/OpenClaw

→

Build
one workflow

→

Use
safely

Plain English version: an AI agent is a helper that can use tools. It can read, write, search, run commands, and remember context — but it still needs clear instructions and human review.

IntroToAI.ppt - Website Preview

Intro to AI: slide preview

Preview the public AI Help starter deck inside the website. Use the arrows to move through the slides, or download the PDF for easy sharing.

Tip: keep this as a safe starter guide: learn the basics, test one boring repeat task, and keep human review on anything that affects real people.

Guide 1 - What is an AI agent?

What is an AI Agent?

Short answer: an AI agent is an AI assistant that can use tools to get work done, not just answer questions.

You ask
a goal

→

Agent plans
steps

→

Agent uses
tools

→

You review
result

Normal chatbot vs AI agent

Chatbot

Answers questions.
Mostly text in / text out.
You do the clicking, copying, saving, and checking.

AI Agent

Can break a goal into steps.
Can use tools like files, browser, terminal, search, memory, and calendars.
Can keep working until a task is complete, then show evidence.

What can an AI agent do?

Pixel mini Michael for 📄 <b>Documents</b>

📄 DocumentsSummarise, rewrite, make checklists, create drafts.

Pixel mini Michael for 🔎 <b>Research</b>

🔎 ResearchSearch, compare options, build a simple report.

🧾 AdminPrepare emails, follow-up lists, meeting notes, SOPs.

🧠 MemoryRemember preferences, projects, and reusable workflows.

🛠️ ToolsRead files, create files, check websites, run safe commands.

Pixel mini Michael for 📊 <b>Business</b>

📊 BusinessSurface exceptions, reduce duplicate entry, help with reporting.

Real workflow examples

Email → CRM → Welcome → Follow-up

Read a new enquiry email, extract name/contact/service/suburb, check for duplicates, enter the lead into a CRM, send an approved welcome email, schedule follow-ups until the customer replies, then hand the conversation to a human with a clean summary.

Quote prep assistant

Read job notes, photos, customer requirements, and pricing rules; prepare a draft quote pack; flag missing information; notify Telegram for approval; only send after human sign-off.

Operations exception radar

Check dashboards and inboxes each morning, find stale leads, overdue jobs, missing payments, or review risks, then post a ranked action list with links and evidence.

Knowledge-base maintainer

Watch repeated staff questions, draft SOP updates, link source documents, ask for approval, and publish a cleaner internal answer so the next person does not ask again.

Customer journey coordinator

Track first touch, booking, quote, job completion, invoice, review request, and warranty follow-up so nothing falls between sales, admin, and field teams.

Personal life admin

Summarise bills, renewals, travel plans, reminders, emails, and documents into one Telegram briefing with tasks and due dates.

The power is the chain: read context → update systems → communicate → schedule next action → monitor for reply → escalate to a human when judgment or trust matters.

What it should not do without you

Spend money, delete files, send public messages, or change business systems without approval.
Make legal, medical, financial, safety, or employment decisions on its own.
Replace human judgment where customer trust, risk, quality, or approvals matter.

First proven prompt to try

I am new to AI agents. Explain what you can do in 5 bullet points, then ask me one simple question so we can try a safe first task.

Guide 3 - Your first safe AI workflow

Your First Safe AI Workflow

A beginner tutorial for turning one annoying repeat task into a reusable AI playbook.

Mini Michael casting a prompt spell for a first AI workflow tutorial

Goal: do one useful workflow in about 30 minutes without giving the AI dangerous access. You will pick a low-risk task, give it examples, check the answer, then save the final prompt so you can reuse it.

Step 1 — choose a boring, safe task

Good first tasks

Mini Michael working calmly on a laptop for safe first AI tasks

Summarise meeting notes, rewrite an email draft, turn messy notes into a checklist, compare options, or make a simple SOP.

Avoid on day oneCustomer databases, payments, deleting files, public posting, medical/legal/financial decisions, or anything that can hurt trust if wrong.

Use fake or copied dataPaste a small sample into the chat. Remove passwords, private customer details, and anything you would not put in a training document.

Define successWrite one sentence: “I want a clean checklist I can follow tomorrow” or “I want a polite email draft I can edit before sending.”

Step 2 — use this starter prompt

You are helping me build one safe repeatable workflow.
Task: [describe the boring task]
Input: [paste notes, transcript, email draft, or requirements]
Output format: [checklist / email draft / table / SOP]
Rules:
- Ask questions only if something critical is missing.
- Do not invent facts.
- Mark uncertain items as [check].
- Do not send, delete, buy, publish, or change systems.
- Give me a short final review checklist.

Step 3 — review like an operator, not a spectator

Accuracy passCheck names, dates, numbers, promises, and any claim the AI may have guessed.

Tone passMake it sound like you. Remove corporate sludge, hype, over-promising, or fake confidence.

Risk pass

Mini Michael facing a robot captcha as a reminder to check AI output

Anything involving money, customers, safety, privacy, or reputation needs a human approval step before action.

Evidence passAsk: “What source line supports each important point?” If it cannot point back to the input, mark it as [check].

Step 4 — turn the good version into a reusable playbook

Once the result is useful, save the prompt plus the expected output format. That is your first workflow, not just one lucky chat.

Workflow name: [Example: Weekly meeting notes to action list]
When to use it: [Every Friday after team meeting]
Inputs needed: [Transcript, decisions, open questions]
Output wanted: [Owner / task / due date / risk / next action]
Human approval required before: [sending to team, updating CRM, promising dates]
Final prompt: [paste the cleaned prompt here]

Step 5 — add the smallest automation later

Manual firstRun it by copy/paste 3 times. Fix the prompt each time. Manual repetition exposes the weird edge cases.

Then semi-automaticLet an agent read one safe folder or draft one message, but keep approval before it changes anything.

Approval gate

Mini Michael holding an approval stamp for human review gates

The magic words are: “draft only, wait for my approval.” Keep that rule until the workflow is boring and proven.

Evidence ruleFor every completed run, keep the input, output, review notes, and what changed. No evidence, no trust.

First workflow ideas

Inbox triage: paste 10 emails → get urgency, owner, next action, and draft replies.
Quote prep: paste job notes → get missing info, risks, and a customer-friendly summary.
Meeting cleanup: paste transcript → get decisions, tasks, blockers, and follow-up message.
Research brief: paste links/notes → get pros, cons, unknowns, and recommendation.

Rule of thumb: if you would trust a careful junior assistant to draft it but still want to review before sending, it is a good first AI workflow.

Guide 2 - Install your first AI agent

Install Your First AI Agent

A complete beginner install path for Mac, Windows, or Linux.

This page is now a walkthrough, not a reference dump.
Pick one agent, pick one computer, then follow the visible path from Step 1 to Step 8. Do not mix Mac, Windows, Linux, Hermes, and OpenClaw commands. If a step fails, stop at that step and use the symptom table near the bottom.

Cost + tokens intro, checked May 2026: Tokens are chunks of text sent in and generated back. Long chats, big files, screenshots, and repeated tool use burn tokens quickly. GPT / ChatGPT OAuth is usually the easiest beginner option when a wizard offers it. Start with ChatGPT Plus; move to Pro only after real rate-limit pain. Provider prices and limits change, so check the provider page before relying on this for business-critical work.

What success looks like

The command works in your Terminal.
You connect one model login, preferably GPT / ChatGPT OAuth if offered.
You run one local hello test before Telegram.
You create one private Telegram bot with BotFather.
The gateway runs and your bot replies to one safe test prompt.
You run doctor/status checks.
You open the dashboard/Web UI last.

What you are about to install

1. Agent app

Hermes or OpenClaw is the local program on your computer. It coordinates tools, memory, messages, and model calls.

2. Model login

The model is the AI brain. OAuth means browser login. API keys are secret passwords. Keep both private.

3. Telegram gateway

The gateway is the bridge between your private Telegram bot and the local agent running on your computer.

4. Dashboard/Web UI

The dashboard is the cockpit. It is useful after install/config checks pass, but it is not the first setup step.

Step 0 — choose one path

Important: this is the only choice section. After this, follow the visible selected panel from top to bottom.

1. Choose the agent

2. Choose your computer

Showing now: Hermes Agent on Mac. Only this computer path is visible below.

Step 1 — before touching Terminal: get these ready

ChatGPT / provider loginHave your account ready. Choose GPT / ChatGPT OAuth in the wizard if offered.

Telegram on phoneInstall Telegram and log in. Your phone is the easiest place to create/test the bot.

BotFather tokenUse verified @BotFather, send /newbot, choose a display name, then choose a username ending in bot. Copy the token privately.

Telegram user IDMessage @userinfobot or @get_id_bot and copy your numeric ID. This lets you allow only yourself.

Safe test folderCreate a folder called AI Agent Test. Put one harmless file inside called test-note.txt.

Permission ruleDay one: no customer data, money, deleting files, public messages, business systems, or full Desktop/Documents access.

In plain English: you are preparing the account, the private message channel, and a harmless sandbox before any installer touches your computer.

Step 2 — selected install path

Follow only the selected panel below. Each panel is a complete install + initial setup path. The dashboard/Web UI appears at the end on purpose.

Hermes required terminal flow

Install prerequisite Git/curl if needed → run the official installer → reopen shell → hermes setup → choose model/OAuth → local hello test → hermes gateway setup → hermes gateway run → hermes doctor / hermes status → dashboard last.

Hermes Agent on Mac — complete beginner path

1. Open Terminal

Press Command + Space, type Terminal, press Enter.

In plain English: Terminal is where you paste setup commands.

2. Check Git

git --version

If Mac asks to install command-line tools, allow it, then run the command again.

3. Install Hermes

curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash
exec $SHELL -l
hermes --version

Success: you see a Hermes version instead of command not found.

4. Run first setup

hermes setup

Choose a local/default workspace. When model/provider is offered, choose GPT / ChatGPT OAuth if available.

5. Confirm or change model

hermes model

In plain English: the model is the brain. OAuth is easier because you log in through a browser instead of copying secret API keys.

6. Run one local chat before Telegram

hermes chat -q "Say hello in one sentence."
# If your version opens interactive chat instead:
hermes

Do not continue until you get a local reply.

7. Connect Telegram gateway

hermes gateway setup
hermes gateway run

Paste your BotFather token and numeric Telegram user ID only into the local setup prompt. Keep this Terminal open for the first test. Stop with Ctrl + C.

8. Verify, then dashboard last

hermes doctor
hermes status
hermes dashboard --host 127.0.0.1 --port 9119 --no-open

Open the shown local dashboard URL only after doctor/status look healthy.

Step 3 — BotFather and Telegram details

In Telegram, search for verified @BotFather.
Send /newbot.
Choose a display name, for example Mike Test Agent.
Choose a username ending in bot, for example mike_test_agent_bot.
Copy the token into a private password note. The token is a password.
Open your new bot and press Start or send /start.
Get your numeric ID from @userinfobot or @get_id_bot. Never send those helper bots your token.
When Hermes/OpenClaw asks for allowed users, paste only your numeric Telegram ID.

If you accidentally share your token: emergency rotation

Open BotFather.
Select your bot.
Revoke/regenerate the token.
Update Hermes/OpenClaw gateway config.
Restart the gateway.

Step 4 — first safe tests

Test 1: can it reply?

Say hello in one sentence.

Test 2: can it explain itself?

Explain what you can do safely. Do not use tools yet.

Test 3: can it respect a folder boundary?

Look only inside my AI Agent Test folder. Tell me what files you see. Do not edit anything.

Permission message to reuse

Before using tools, show me a short plan. Do not delete files, send messages, spend money, or touch business/customer/private data.

Step 5 — troubleshooting by symptom

What you see	Probably means	Try this
`command not found`	Terminal cannot find the app yet.	Close/reopen Terminal, run `exec $SHELL -l`, then check version.
`permission denied`	The command lacks permission or wrong folder.	Stop. Do not randomly add `sudo`. Re-read that step.
Model says no auth	Provider login/API key not connected.	Re-run `hermes model` or `openclaw configure --section model`.
Telegram bot silent	Gateway off, token wrong, bot not started, or user ID not allowed.	Send `/start`, check token/user ID, then run gateway foreground again.
Dashboard blank or useless	Dashboard opened before setup/gateway/doctor was healthy.	Go back to doctor/status checks. Dashboard is last.
Windows path feels broken	PowerShell route hit Windows-specific issues.	Use the WSL Ubuntu route in the selected panel.
OpenClaw complains about Node	Manual runtime mismatch.	Use official installer, or update to Node 24 / at least Node 22.14+ before npm fallback.

Step 6 — how to stop safely

Foreground gateway: press Ctrl + C in the Terminal running it.
If Telegram feels wrong: stop gateway, remove bot from groups, rotate token in BotFather.
If unsure: stop. Do not connect customer data, payment tools, or public posting.

Step 7 — maintenance and dashboard reminder

Monthly health checkRun update/doctor/status, send one safe prompt, and check provider usage/cost.

hermes update
hermes doctor
# or
openclaw update
openclaw doctor

Dashboard/Web UI ruleOpen it only after install, model login, Telegram gateway, and doctor/status are healthy. Keep it local while learning: 127.0.0.1, not public internet.

Contact

Send me the messy workflow.

I am most useful where a business has scattered systems, repeated admin, slow handoffs, unclear reporting, or AI ideas that need safe review gates before they touch customers, money, or operations.

AI workflow strategyService-business automationOperational visibilityAgent review systemsCRM / ops handoffsDigital infrastructure

Email Michael

Recycle Bin

Nothing to delete. Unreleased apps are safely hidden.