We are an agency, not a software vendor. But the reason a small team can run your web, search and ads together — and show you a record of every change — is that we wrote the software we run them with.
You get named people, not an account-manager layer in front of the people doing the work. The person who approves a change is someone you can email.
Each agent has one job — mine search terms, pace budget, audit tracking, flag stale pages — and a shift it runs on. They inspect and propose; they cannot apply.
Web changes and ad changes land in the same place, with the evidence attached. That is why the two sides of your marketing stay in step.
Every applied change carries what, why, who approved it and when. If you leave, the accounts, the content and the history are yours.
Two platforms we wrote and run ourselves. They are how the work gets done faster and gets recorded properly — not a product you have to learn. Most clients never log in.
Pages, locations, forms and revisions live in one system of record. A crew of agents each own one job — freshness, internal links, structured data, location pages, metadata — and run it on a schedule. Every edit to a live page is a revision waiting for an editor.
Google, Microsoft, Meta and ChatGPT in one place. Agents mine search terms, pace budget, audit conversion tracking and rewrite tired copy around the clock. Everything they propose lands in a single approvals queue, and sold-lead outcomes go back to the platforms.
You are not buying software. Lumin CMS and Lumin Ads are internal operating infrastructure. You buy web, search and advertising as a service; these are what let a small team run them at the pace of a much larger one, and leave a record of every change.
See how it worksNothing consequential reaches a live account or a live page on its own. Four roles, in order, every time.
The approvals queue in Lumin Ads, running on a demo account.
What the gate looks like from your side: no ticket queue, no Tuesday meeting. An agent reads the live account, a person decides.
“We're still getting calls about the $300/day ad. Can we fix it?”
Reads the live account. Finds the price extension. Drafts the removal.
Checks it, approves it. Applied. You get a reply, not a ticket.
An illustration of the approval path, not a response-time guarantee. Turnaround depends on the change and who has to sign it.
Your own app, in your brand. Push notification, tap to call, speed-to-lead measured. One plain-English email every Friday. No dashboards.
A property of the CMS, not a service-level promise. The publish is the deploy.
Agents have no apply permission. Every consequential change carries a person's name.
A proposal over the cap is blocked at the apply gate — even when a human has approved it.
Outcome numbers — leads, cost per qualified lead, revenue — belong to a client and a timeframe, so we publish them on the case study that can carry the baseline and the method, not in a headline.10 client sites live today →
Read. Agents have read access to your accounts and content, and no apply permission. They can inspect, research, draft and propose. Applying anything consequential requires a named approver, and the spend caps are enforced separately at the apply gate.
The agents are driven through Claude and local harnesses. Which model runs a given job is an implementation detail we change when a better one ships; the approval gate and the caps do not change with it.
It gets rejected, and the rejection is part of the record. A proposal that exceeds a hard cap never reaches an approver's judgement at all — it is blocked at the apply gate.
Yes. Clients who want visibility get access to the approvals history for their own account. Most prefer the weekly email, so that is the default.
No. Most clients never log in. The deliverable is the marketing, not the tool.
Website, local and AI search, ad account, or all three. Read-only access, real findings on your own accounts, and a person who walks you through them. You keep the findings either way.