Make Money With Hermes Computer Use (Service Stack 2026)

Hermes Computer Use is the highest-ROI free install I've put into my agency stack in 2026, and a one-person operator can now run agency-grade services on top of it at near-zero marginal cost. The agent operates your Mac in the background — clicks, types, scrolls, drags — while you sell, fulfil, and onboard clients on the same machine.

This article is the money view of Hermes Computer Use. I'll cover the services you can sell off the back of it, the pricing model that works, the install, the prompting framework, the live tests that prove it earns its place in a service stack, and the routine you can ship to clients next week.

🔥 Want my full Hermes Computer Use guide + 100 prompts? AI Profit Boardroom has the 100-prompt guide, 30-day roadmap, step-by-step SOPs, daily custom-video Q&A, and 4 weekly coaching calls with 2,900+ entrepreneurs. → Get inside

Why Hermes Computer Use Is A Money-Making Tool

Service businesses get bottlenecked on operator hours. Whether you run an SEO agency, a VA service, an ops consultancy, or a content shop, the cap on how much you earn is the cap on how many operator hours you can buy or sell.

Hermes Computer Use breaks that cap. The agent operates a Mac in the background while you do other work on the same machine. One operator suddenly has the throughput of three. Your gross margin per client goes up because your fulfilment cost per client goes down.

The compounding maths is unreal. If you save sixty minutes per client per week across a portfolio of twenty clients, that's twenty hours of operator capacity recovered. At a billable rate of £150/hour, that's £3,000 per week of recovered or resold time. For a free install.

What Hermes Computer Use Actually Is

Hermes Computer Use is a brand new free update to the Hermes Agent framework from Nous Research. It lets the agent operate your Mac the way you would. It reads the screen, identifies every button, picks the right one, clicks, types, and chains the next step.

All of this happens in the background. Your cursor doesn't move. Your focus doesn't change. The agent runs underneath your active workflow rather than on top of it. You and the agent share the same Mac as a two-operator team would share an office.

It uses an MCP server under the hood. It's currently Mac-only. It's free and open source. It works with any vision-capable AI model — Claude, GPT, Gemini, OpenRouter, or a free local model.

Eight Services You Can Sell On Top Of Hermes Computer Use

I'll skip the theory and go straight to the money. Here are eight productised services a one-person operator can ship on top of Hermes Computer Use this month.

Inbox triage as a service. £500-£1,500/month per client. You run Hermes in the morning to triage their inbox, draft replies for the urgent emails, and leave them in Drafts for approval.

Content capture and second brain build. £1,000-£3,000/month per client. Hermes captures voice notes, transcribes them, organises them in Obsidian, links to relevant projects, builds the client's second brain over time.

File and folder organisation. £300-£800/month per client. Hermes runs nightly to organise the client's Desktop, Downloads, and project folders into a consistent structure.

Meeting follow-up automation. £500-£1,500/month per client. After every client call, Hermes extracts action items, pastes them into their project tracker, drafts recap emails, and creates follow-up tasks.

Personal knowledge management. £800-£2,500/month per client. Hermes captures the client's daily lessons, links them in Obsidian, builds a searchable knowledge base.

Research and brief preparation. £1,500-£4,000/month per client. Hermes opens browser, gathers source articles, summarises them, drops the synthesis into the client's project folder.

Cross-app workflow automation. £1,000-£3,500/month per client. Hermes runs the cross-app shuffles that used to eat the client's afternoon — meeting notes to project tracker, voice notes to Obsidian, draft emails to Drafts folder.

Agent-to-agent pipeline service. £2,500-£7,500/month per client. The premium tier. Multi-agent pipelines that hand off work to each other while the client is on calls or focused on strategic work.

Stack any three of these for a single client and you're at five figures monthly recurring revenue from one human operator who runs Hermes Computer Use in the background.

The Pricing Model That Works

Three pricing models that actually convert in my experience.

Done-for-you monthly retainer. £1,000-£5,000 per client per month for one service. £3,000-£10,000 per client per month for a stack of three. This is the simplest model and the one most agencies should start with.

Outcome-based pricing. Charge per hour of operator time recovered for the client. Quantify the saving, take a percentage of the value, ship the bill. Higher upside if your client values their time at £200/hour or more.

Productised setup plus maintenance. £2,000-£5,000 upfront for the setup (install, configure Hermes, build the prompts, train the client) plus £500-£1,500/month maintenance. This is the model I'd recommend for SEO and marketing agencies adding Hermes services to their existing stack.

The model that doesn't work is hourly billing. You're selling AI leverage. Don't sell yourself by the hour.

Installation In One Command

The install is genuinely a single command. Open your terminal and type hermes computer-use install. Press enter. Wait under a minute. Done.

You'll then need to grant Accessibility permissions in System Settings under Privacy and Security. Allow Terminal (or whichever shell you launched Hermes from). Without this step, the agent can't see the screen and the rest of the workflow falls over.

Restart Hermes once. The Computer Use toolset shows up ready to use. Validate with a basic prompt like "open Notes" — if it pops Notes open in the background, you're live and ready to sell.

For agencies, this means client onboarding takes minutes, not days. You can install this on a client's Mac during the first kickoff call.

The Prompting Framework — Goal, Oversee, Stack, Transform

The framework I teach the people I sell Hermes services to.

Goal. Hand the agent a complete client outcome rather than micro-tasks. Bad: "click new note." Good: "read today's voice notes, transcribe them, drop summaries into the client's Obsidian, tag them by project."

Oversee. Watch the run. Step in only if something looks off. Don't micromanage. The client is paying you to run the system, not to operate every click.

Stack. Different steps can use different models. Reasoning-heavy steps go to Claude. High-volume routine steps go to a free OpenRouter model. Privacy-sensitive steps go to a local model.

Transform. Every run produces a real artefact the client can see and value. Drafts in their inbox. Notes in their Obsidian. Files in their folders. The output is the receipt.

Live Test 1 — Background App Open

The simplest test. Prompt: "open Notes." Hermes opened the app in the background. Cursor didn't move. Focus didn't change. App appeared in seconds.

This is the smoke test for any client install. If you can't pass this on day one, the rest of the service doesn't work. Hermes passed cleanly.

Live Test 2 — Two-Skill Personalised Note

Bigger test. Prompt: "Open Notes and create a new note journaling about the best ways you could help me save time day-to-day."

Hermes stacked Apple Notes plus Mac OS Computer Use and produced a complete personalised note with ten concrete ideas in seconds. Inbox triage. Content research. AI SEO context. Personal knowledge capture. Second brain surfacing. The ideas read like a thoughtful colleague's notes, not generic AI filler.

This is the test that proves you can demo Hermes to a client in the first ten minutes of a sales call and close them. They see real output, on their own machine, in seconds.

Live Test 3 — Where It Hits Its Limit

I tried something bigger. Prompt: "Go into Obsidian, organise my knowledge base, add details and context, add emojis and titles, organise folders, improve the knowledge graph."

The good first sign was that Hermes asked permission before any destructive move. The guardrails worked — important if your client is paying you to run agents on their machine.

The honest finding is that long multi-step tasks are slower than expected right now. I stopped it after a few minutes. The agency lesson is to scope client services to three-to-five step workflows and chain shorter pipelines together rather than promising one giant pipeline.

Live Test 4 — Agent-To-Agent Service

The meta test. Codex hit a token limit so I switched to Kimi K2.6. I prompted Hermes to open a new terminal, start another Hermes inside it, and say hello.

The first Hermes opened the new terminal. It started the second Hermes. It typed "hello." The second Hermes responded: "Hey Julian, what are we working on today?"

Two agents talking on the same Mac. The premium service tier I mentioned earlier — agent-to-agent pipelines — is built on exactly this primitive. It works today.

Service Stack Walkthrough

The walkthrough above shows the real Hermes Agent setup that powers the service stack — model selection, workflow design, and the patterns I've found most reliable for client work.

Best Models For Agency Use

The agent works with any vision-capable model. Text-only models won't work because screenshots are how the system identifies buttons.

Claude. My default for reasoning-heavy client work — strategy notes, complex email drafts, anything that needs nuance.

OpenRouter. Useful for agencies because you get 200+ models behind one API key. Bill the client at a slight markup on the API costs.

OpenAI GPT-5.4 or Codex. Solid for general client work. Codex burns tokens fast, so use it for short jobs.

Local models — Gemma 4 via Ollama or LM Studio. Essential when the client has data that can't leave their machine. This is your enterprise-tier offer.

Skip text-only models. No vision, no computer use.

Token Economics For Agencies

Computer-use sessions take a screenshot at every step. Token usage adds up fast. Codex hit its limit during my testing.

The agency move is to route high-volume client work through free APIs and bill the client for premium model usage only on the steps that genuinely need it. Step 3.5 Flash on Nous Portal is currently free and fast enough for most computer-use moves. OpenRouter has free models too.

Your gross margin on Hermes services improves dramatically when you nail the model routing. Charge the client a flat retainer. Run 80% of the work on free APIs. Pocket the difference.

Comparison Table — Hermes Computer Use Vs Other Service Layers

Tool	Touches native Mac apps	Background mode	Free	Permission guardrails	Agency fit
Hermes Computer Use	Yes	Yes	Yes	Yes	One-person ops
OpenClaw	Yes	Partial	Yes	Weaker	Browser-heavy services
Manus	No	No	Paid	Yes	Cloud-only services
Operator (OpenAI)	No	No	Paid	Yes	Web automation services
Zapier	No (API only)	N/A	Limited free	N/A	Integration services

For agencies selling background Mac workflows, Hermes is the only option that does all four things — runs in the background, costs nothing to install, ships proper guardrails, and is genuinely native on macOS.

🚀 Free SEO Strategy Session — Goldie Agency Want to pair Hermes Computer Use with real link building? Book a free strategy session with my 7-figure SEO agency (50-person team). → Book free session

Safety, Client Trust, And Permission Hygiene

Selling agent services to clients lives or dies on trust. The first time an agent goes off the rails on a client's machine, you lose the account.

Hermes Computer Use ships with multi-layer guardrails. Anything destructive — deleting files, sending emails, moving documents permanently — requires explicit permission from you (or your client) before the agent acts. This is the design choice that makes agency-grade selling possible.

Three rules I follow on every client install.

First, never point the agent at the client's billing dashboard, banking app, or production system on day one. Build trust on low-stakes tasks. Graduate slowly.

Second, log everything. Every run leaves an audit trail. Show the client the logs in weekly reviews. Transparency is the differentiator.

Third, build a clear stop-condition into every pipeline. No infinite loops. No agents running unattended for hours. Bounded runs only.

Honest Limitations For Agency Use

Three things to be clear-eyed about before you start selling.

Complex tasks are slower than basic ones. The Obsidian reorg from test three was at the edge of reliable today. Scope client services to three-to-five step workflows.

Token weight on premium models. High-volume client work on Claude or Codex can rack up costs. Route to free APIs for routine work and bill clients accordingly.

Mac-only. If your client is on Windows or Linux, this isn't ready for them yet. Cross-platform support will come.

None of these stop you from launching a Hermes service stack today. They're scope notes.

Agency Routine I Run Daily

The routine I run on my agency stack. Total active operator time per client per day: under ten minutes. Margins per client: high.

6am. Inbox triage pipeline runs for each retainer client. By 9am the day's emails have draft replies in their Drafts folder.

9am. Content capture pipeline runs while clients are on their first calls. Voice notes get transcribed and tagged into their Obsidian.

12pm. File organisation pipeline runs over lunch on every retainer client's Desktop and Downloads.

3pm. Meeting follow-up pipeline runs after client call blocks. Action items in their tracker. Recap emails in their Drafts.

9pm. Agent-to-agent research pipeline runs overnight for clients on the premium tier. By morning they've got synthesised position notes ready for their day.

One operator. Multiple clients in parallel. Background mode means no operator context switching. Margins north of 80% on a productised service.

Belief Shifts For Service Sellers

Four lies I hear from service operators still on the fence.

"AI services don't sell." False. They sell faster than any service category I've launched in the last five years.

"Clients won't trust an AI to touch their Mac." False if you sell the safety story properly. The permission guardrails and audit trail are the trust mechanism.

"I'll need a team to deliver this." False. One operator can run twenty retainer clients on Hermes Computer Use with the right routine.

"I'm too late, the market is saturated." Completely wrong. The feature dropped this week. Almost no agency has built a productised service on it yet. The window is wide open.

FAQ — Make Money With Hermes Computer Use

Can I really sell services on top of Hermes Computer Use?

Yes. The eight productised services in the section above are all viable today, with retainer pricing from £300 to £7,500/month per client.

Is Hermes Computer Use actually free?

Yes. The tool is free and open source. You only pay for the AI model API calls, and those can be routed through free options like Step 3.5 Flash for most agency work.

Does it work on Windows for client work?

Not yet. Mac-only currently. Build your agency around Mac-based clients first.

How do I install it on a client's Mac?

One command: hermes computer-use install. Then grant Accessibility permissions in System Settings. Five minutes during a kickoff call.

What if my client hasn't installed Hermes Agent yet?

Start with my Hermes Agent Installation Guide first.

Will the agent disrupt the client's work?

No. Background mode means their cursor doesn't move and their focus doesn't change while the agent runs.

Which model should I use for client work?

Free OpenRouter models or Step 3.5 Flash for routine work. Claude for reasoning-heavy steps. Always disclose the model stack to clients so there are no surprises.

Can Hermes delete client files without asking?

No. Destructive actions require explicit permission. This is the single most important safety guarantee for selling agent services.

How does it compare to OpenClaw for agency use?

Better guardrails and cleaner background behaviour. See OpenClaw Computer Use for the side-by-side.

Should I join AI Profit Boardroom?

If you want the 100-prompt guide, the 30-day roadmap, the step-by-step SOPs, and the weekly coaching calls that walk through agency Hermes services on screen-share — yes. The 7-day refund makes it risk-free.

Latest Updates

Hermes Agent Goals (Persistent Autonomous Loops) — the autonomous loop layer that powers premium-tier services.
Hermes Agent HUD UI — visual control panel for monitoring client runs.
Hermes MCP Server — the protocol layer that powers the service stack.