The engine runs 38 commands across eight modules: grading your agency, auditing every page, measuring AI search visibility, writing blogs and service pages, social calendars, Google Business Profile, Google and Meta ads, and full outbound. It runs on your own machine and writes files you own.
What it does
Thirty-eight commands, inbound and outbound.
It runs inside Claude Code on your own machine or VPS. You type a command, it does the work, and it writes files you own. Nothing publishes without your approval.
Grade your agency
| Command | What it does |
|---|---|
/agency-report | Grades what your agency delivered across six dimensions, and divides what you paid by what measurably changed |
/grade-content | Scores the next deliverable before it goes live, and measures whether the same article went to the agency's other clients |
/audit-site | Every page, plus the cross-page problems a single-page audit cannot see |
/audit-page | One page, one ranked fix list, about a minute |
/competitor-gap | What the firms outranking you have that you do not, ordered by impact over effort |
AI search
| Command | What it does |
|---|---|
/ai-visibility | Share of voice across 25 fixed buyer prompts, tracked month over month, plus the domains being cited instead of you |
/schema | Organization, LocalBusiness, Service, FAQ and Breadcrumb JSON-LD, generated and validated |
Produce the work
| Command | What it does |
|---|---|
/buyer-questions | The real questions buyers ask, mined from People Also Ask, Reddit, Quora and review sites |
/write-blog | Full top-ten SERP and question research, then a cited draft that passes the gates |
/write-page | Service and location pages |
/refresh-content | Rewrites a page that is not performing, against what now ranks |
/social-calendar | A dated quarter of LinkedIn and Google Business posts, then it writes every one |
/schedule | Queue, approve once, publish on the times you set |
Local
| Command | What it does |
|---|---|
/gbp-audit | Scores your Google Business Profile across eight dimensions and hands you every change it needs, including the ones that risk a suspension |
/gbp-post | Posts that feed the local pack, where near-me searches resolve |
/local-audit | NAP consistency between the site, the schema and the listing |
Ads
| Command | What it does |
|---|---|
/negative-keywords | Mines your search terms and lists every query burning money. Works from a CSV export, no API approval needed |
/google-ads-report | Wasted spend as a dollar figure, plus the structural problems to raise with whoever runs the account |
/meta-ad-copy · /meta-ad-creative | Primary text, headlines and branded images in testable variants |
/meta-ads-report | Performance, creative fatigue, and the specific changes to make |
Outbound
Inbound waits for a buyer who is already looking. This goes and finds the ones who are not looking yet, which for most MSPs is where the next ten clients are. Runs about $200 a month for both channels.
| Command | What it does |
|---|---|
/icp | Your ideal customer as machine-readable gates, not a description. Disqualified accounts get blocked automatically |
/find-prospects | A tiered list built off real trigger events |
/research-account | A cited brief on one account before you contact them |
/write-sequence · /linkedin-dm | Multi-touch email sequences and LinkedIn openers, validated for structure and tone |
/handle-reply · /meeting-prep | Objection handling and the one-page brief before the call |
/deliverability | Checks the sending domain before a campaign burns it |
The guardrails are the product
Every output passes the same gates before it is called done. A statistic with no real source URL is blocked. A certification not on your verified list is blocked. "100% secure" and "guaranteed protection" are blocked. Structured data that does not parse is blocked.
You sell trust. One fabricated breach statistic on your site costs more than this did. These gates are why you can let it write without reading every word hunting for an invented badge.
What running it actually looks like
Setup is one command. The engine interviews you for about twenty minutes, covering your services, service area, competitors, and the credentials you will stand behind, then validates every connection and runs your first audit, AI visibility check and agency scorecard without being asked.
After that the rhythm is monthly rather than daily. Run the scorecard before the agency review. Grade the next deliverable when it lands. Produce a quarter of social in one sitting, approve it once, and let the scheduler post it. Most owners spend an hour or two a month in it and the rest of the time it is simply the thing that answers questions when they come up.
What it will not do
It will not publish anything without your approval, and there is no setting to change that. It will not guarantee a ranking, a citation or a lead, because nobody can. It will not replace judgement, a point of view, or someone who picks up the phone when a client says something worth writing about.
It also will not change a live Google Business Profile or an ad budget on its own. Those are reversible only at some cost, so it recommends and a human clicks.
Do I need to be technical to run this?
You run an IT company. If your team can deploy software on a VPS and manage an API key, this sits well inside what you already do. Setup is one command and a conversation.
Can I use it for my clients as well as my own marketing?
Yes. The licence covers your own marketing and your clients' marketing. It does not permit reselling or redistributing the engine itself.
What happens when Claude Code or an API changes?
Fixes are pushed to every live seat for twelve months from purchase. After that the release keeps working as it is, and updates move to an hourly rate.
Does it need Ahrefs or Semrush?
No. No report in the engine depends on a keyword suite. Connect one if you already pay for it, but it is not required and nothing is degraded without it.
How much does it cost to run each month?
Between $50 and $260 for the core, covering a Claude subscription, Firecrawl and a small VPS. Add $100 to $300 for a sending platform and contact data only if you run outbound.