The Omnia Agent: Your AI Visibility Teammate, Live Today

A few months ago we shipped the Omnia MCP. The pitch was simple: connect Omnia to Claude or ChatGPT, and your AI visibility data stops living in a dashboard you have to remember to open. You ask your assistant a question, it reads your share of voice, your citations, your sentiment, and it can even set up monitoring or generate insights for you. And customers found that incredibly useful.

But we watched what happened next, and it taught us something we did not fully expect.

The MCP gives a general assistant access to your data. What it does not give it is a method. Claude will happily read your citation data and write you a plan, but it does not know that a tactic already failed for your brand last month. It does not know which of your pages are losing the citation and why, unless you re-explain the whole situation every time. Every conversation starts from zero. The intelligence is there. The context, the memory, and the accumulated judgment about your specific brand are not.

That gap is the reason the Omnia Agent was born.

What the Omnia Agent actually is

The Omnia Agent lives inside Omnia and does your AEO work. Not a chatbot, and not a general assistant we point at your account. It is closer to hiring an AEO specialist who has been with your brand a while: it knows your AI visibility history, and works one move at a time. It earns the right to act by being right often enough that you stop double-checking every step.

The job it works toward is narrow and stable: improve how your brand shows up in AI answers. Everything else follows from that. It reads what engines like ChatGPT, Claude, and Perplexity say about you, decides what is worth chasing, creates or rewrites the content, places it, and then measures whether the move actually worked. Then it does it again, informed by what it just learned.

The most important part: it acts, but not behind your back. Reading, drafting, and proposing run freely, because those are reversible and cost you nothing. Anything that publishes, sends an email, spends credits, or deletes data waits for your explicit sign-off. This is not a marketing line. It is enforced in the product, per class of action, and you can tune it. If you want your Omnia Agent to stop asking about credit spends under five euros, you tell it that in chat, and it remembers.

Why we did not just make the MCP smarter

This is the honest version of the decision, because the alternative was real and we considered it.

We could have kept pushing on the MCP path: better tools, better prompts, let the general assistant carry more of the load. The problem is that a general assistant is general on purpose. Its strength is breadth. AEO does not reward breadth. It rewards a specific method applied consistently to a specific brand over time, with memory of what has already been tried.

So the difference between your Omnia Agent and "Claude with the Omnia MCP connected" is not intelligence. It is everything around the intelligence. Your Omnia Agent carries durable, per-brand state: a living roadmap of what it is chasing and why, a record of every move it has made, and a tail of tactics that were tried and did not move the needle, so it does not waste your credits proposing them again. It connects to where your work actually lives, like Search Console, your CMS, and your inbox.

If you want the direct comparison, we put one up here. We are not saying general assistants are bad at this. We are saying that an employee who remembers is worth more than a brilliant contractor you have to brief from scratch every morning.

The same work, with and without your Omnia Agent

The clearest way to see the difference is to watch the same task move through both worlds.

Without your Omnia Agent, improving your visibility for one prompt looks like this: you open the dashboard and read that you are absent from a prompt your buyers actually ask. You export the citation data. You figure out which third-party pages the engines are pulling from. You audit your own page against the ones getting cited. You brief a writer, or write it yourself. You publish. You remember, three weeks later, to check whether it moved. Usually you don't remember. If it did not work, you have no record of that, so a month later someone on the team suggests the same tactic again.

With your Omnia Agent, the same task is a conversation. You point it at the prompt. It builds the gap analysis from your own monitoring data, audits the page against the cited winners, drafts the rewrite or a new piece, and shows you an outreach pitch to the sites the engines cite. You approve what you want to go live. It records a baseline the moment it proposes the move, sets a date to check back, and on that date it re-reads the metric and tells you whether it worked. If it did not, it files the tactic under "tried, did not move" and does not suggest it again. The context lives in the roadmap, not in your memory.

The honest summary: without your Omnia Agent, monitoring tells you where you stand and the doing falls to whoever has time, which on a small team is usually nobody. With your Omnia Agent, the doing has an owner and you move from deciding every step to approving the ones that matter.

How it works, briefly

Your Omnia Agent runs a loop anchored on a roadmap. The roadmap is the source of truth for what to do next, and it is deliberately small: what is active, what is done and waiting to be measured, what has been measured, and what was tried and did not work.

It works one move at a time rather than firing off everything at once. This is a choice about leverage and about not overwhelming you. Moves are scored roughly by how much intent a prompt carries, how big your gap is against competitors, and how far the relevant engine reaches, divided by the effort involved. The highest-leverage move goes first.

Underneath, each brand's state is a versioned repository, and your Omnia Agent itself runs in its own isolated sandbox that idles to near-nothing when you are not using it and wakes up in about a second when you are. You will never see any of this, and that is the intent. The runtime is supposed to be invisible. What you should see is an agent that remembers, sequences sensibly, and does not touch anything irreversible without asking.

The skills it ships with

An agent is only as good as what it actually knows how to do. The Omnia Agent ships with a bundled set of skills, which are the encoded AEO expertise a specialist would carry in their head. These are maintained by us and read-only, so they get sharper as we learn without you having to do anything. The set covers the full loop:

• Site audit, which reads your site and reports on the technical, SEO, crawlability, and AEO issues that keep engines from citing you.
• Prompt discovery, which finds, scores, and clusters the questions your buyers actually ask engines into a prioritized brief.
• Competitor profiling, which works out who is winning your prompts and where the engines are pulling their answers from.
• Page rewrite, aimed specifically at pages that are losing a citation they used to hold.
• Content creation, which is a first-class module rather than a single skill: brand-new pages for questions you have no page for yet, comparison and listicle pages, non-article assets like video or original data, and valid structured data.
• Placement, which publishes approved content and submits it for faster indexing.
• Entity correction, which fixes the third-party records engines rely on to describe you.
• Outreach, which drafts a pitch to a site the engines cite and, once you approve it, sends it.
• Finally, measurement, which closes the loop by re-reading the metric each move was meant to move.

On top of the bundled set, your Omnia Agent learns skills you teach it in conversation. If you explain how your team likes a certain kind of page written, that becomes a skill it keeps, owned by you, editable and removable, and shared across your team's brand. The bundled skills are the craft it arrives with. The learned ones are how it adapts to your craft specifically.

What you can ask it to do today

Concretely, here is the kind of thing you can say to your Omnia Agent right now and have it act on.

• Ask it to read and make sense of your own data: "Where did my share of voice drop this week, and against whom?" or "Which prompts am I absent from that my competitors own?" It reads the monitoring data directly and answers with your real numbers, not generic advice.
• Ask it to diagnose: "Audit my pricing page against the pages ChatGPT cites for this prompt" or "Run a technical AEO check on my site and tell me what is blocking citations." You can ask it to find work: "Find the high-intent prompts my buyers are asking that I am not tracking yet" and then, with your confirmation, add them to what Omnia monitors.‍
• Ask it to create: "Draft a comparison page for this prompt in the format the engines are citing" or "Rewrite this post so it holds the citation it is losing."
• You can ask it to reach out: "Pitch us to the roundup that dominates citations for this prompt," which it drafts and only sends once you approve. And you can ask it to hold itself accountable: "Report on everything you have done this month and whether it moved the metric."

You can also manage the account itself through conversation: pause or add prompts, generate or dismiss insights, edit what it knows about your brand, set how much it is allowed to spend, and edit its memory of your preferences. The rule underneath all of it is consistent. Reads and drafts happen freely. Anything that publishes, sends, spends, or deletes waits for your yes.

What the Omnia Agent does not do

We think the boundaries are as important as the capabilities, so here they are plainly.

It does not publish content or send outreach without your approval. It does not chase hacks; the roadmap is built from tactics that already work in your category, not from tricks. It does not fabricate signals, which means no fake reviews, no inauthentic mentions, no link buying. When a negative thing an engine says about you is actually true, your Omnia Agent surfaces it to you as a signal about your product rather than trying to optimize it away. And it does not spin content out of thin air. It needs your real substance, your product, your data, your point of view, because that is what engines eventually reward and what fabrication cannot fake.

It also stays on mission. If you ask it for coding help or trivia, it will point you back to your brand's AEO. It is an employee with a job, not a companion.

Where this is honestly at today

The Omnia Agent is live today, and you can use it. Some of what we have described is fully shipped and in daily use: the Omnia Agent, its method, its memory, the connectors to Search Console and Analytics, the approval gating, and content creation. Some of it is defined and coming: a proper task queue, a richer memory browser, a skills library you can teach and share across your team, and the scheduled proactive check-ins where your Omnia Agent surfaces itself when something meaningful moves rather than waiting for you to open it. We would rather tell you where the line is than let the demo imply the line is not there.

We are also still working out pricing. For now it is free during the 14-day trial, which is genuinely expensive for us to run, so that will not last forever. If you have been meaning to take AI visibility seriously but never had the bandwidth or an AEO specialist on the team, this is the cheapest that specialist will ever be.

Point your Omnia Agent at one prompt you want to win. Let it audit the page, draft the fix, and reach out to the third parties the engines cite. Then ask it to report back on whether it worked. That is the whole loop, and it is the fastest way to see whether this is real.

Or start wider: your Omnia Agent can read across all your prompts and recommend a monitoring strategy, so you climb positions gradually instead of chasing them one at a time.

Hire Omnia Agent as your AEO colleague, for free for 14-days.