neo · the agent
Neo, the AI agent dedicated to your brand’s paid media.
Neo is an AI agent that’s an expert in cross-channel paid media, deployed on a machine dedicated to your brand. It knows your brand, reads your accounts, delivers reviews, audits and reporting on schedule to Slack or Gmail, and writes your concepts and briefs on request.
4,000 advertisers and agencies already use Creaboost.
You ask, Neo produces the deliverable.
Here, Neo builds 25 personas for Maison Lune, each with a first concept, saves the document in the brand’s Docs and publishes it with a public link.
what neo does
Neo keeps working when you’re not.
methods
Performance reviews
Every week and every month, a cross-channel business review and a review per platform, with 3 to 5 dated decisions.
skills →Account audits
Structure, tracking, budget and creative: Neo audits an account and ranks what needs fixing, from the costliest issue down.
skills →Creative briefs and concepts
Personas, angles, offers, formats and hooks, drawn from your customers’ own words.
create →
data
market
Competitive watch
Your competitors’ ads, their stores and their customer reviews.
discover →
execution
memory
Brand memory
Brand platform, customer voice, personas, concepts and Lexicon, built at onboarding and kept current.
brands →
your tools
Notifications
Neo reaches you where you already work, in Slack or Gmail. Teams, Outlook, Notion and Google Drive connect the same way.
slack and gmail →Exports
Audits, briefs and reporting export to Google Slides, Sheets and Docs, in your brand’s look.
docs →Shareable pages
Any deliverable can be published as a web page, public for a client or workspace-only, at the same address every time it updates.
docs →
why a dedicated agent
The model reasons brilliantly, as long as it gets your context.
Plugging the Meta MCP into Claude and pasting in your Shopify and GA4 exports almost always ends the same way: long, poorly calibrated analysis that goes straight to the trash. Three gaps explain the slop.
- methods
The rules of 2019 and 2026, mixed up.
Trained on thirty years of content published online, the model gives a 2019 tip as much weight as a 2026 one. It still recommends lookalikes and interest targeting, even though Andromeda rewrote the rules at Meta.
- memory
A brand it meets for the first time.
At best, it sees your website. It knows nothing of your personas, your customers’ own words, the concepts already tested or what worked, so it fills the gap with generic advice like “test new audiences.”
- data
Raw data, in silos.
The Meta MCP gives it a single platform, as raw data, with Meta’s own attribution. Your Shopify and GA4 exports come with no shared definitions, Google, TikTok and organic stay out of view, and thousands of rows flood its working memory.
Neo gives the model what it lacks: dated, verified methods for every channel, a written understanding of the brand, and clean, cross-channel, already-synthesized data. One brain, no silos, reading all your data and delivering only what matters.
A Neo decision cites its number, its rule, its steps and its check.
Here is one decision from the Maison Lune weekly Meta review, as Neo writes it: the number behind it, the rule applied, the steps for your team in Ads Manager, and the date you’ll know whether it worked.
Scheduled tasks →decision 2
Scale back the US campaign.
ROAS 1.07 on 7-day click, with a CPM four times that of France.
rule
A market that stays far from the benchmark after the learning phase loses its budget.
A rule from the meta_pilot skill, which Neo follows in the Meta review.
steps
On Oct 8, cut the budget from $320 to $160 a day.
Once 7 days have passed since the last significant edit. Then rename the two ad sets called “Europe” that target the United States.
check · oct 15
Cut it if the 7-day click ROAS stays under 1.5.
Your team decides and acts: Neo has no write access to your ad accounts.
under the hood
Neo is a frontier model equipped with methods, data, the market and execution.
model
Claude Code, on your team’s Claude subscription.
claude opus 5.5
methods
One skill per deliverable: what to look at, and how to judge it.
data
Every source you connect: Meta and Shopify read live, the others synced several times a day.
market
Your competitors’ ads, their stores and their customer reviews.
execution
Scheduled tasks that deliver reviews, alerts and reporting to Slack or Gmail.
neoAn agent dedicated to your brand, steering by concept, ad set and creative surface.
setup
Neo starts on your brand in minutes.
01
Connect your Claude subscription.
Neo runs on Claude Code with your team’s subscription. Your Claude credentials never go through Creaboost.
02
Connect your ad platforms and your store.
Meta, Google, TikTok, Shopify and your other sources. If you don’t have access, send a simple link to whoever does. The tokens stay in an isolated service, out of the agent’s reach.
03
Pick your tools and your reviews.
Slack, Gmail, Google Drive or Notion connect as Claude connectors. Drop in your files (brand book, studies, price list) and keep the proposed review setup: one review per platform every Monday, then the business review.
04
Neo builds the brand memory.
The Brand Kit first, then products and customer voice, economics, the brand platform, personas, events, and concepts for your three priority personas. Its questions come all together at the end, each with a recommendation.
05
The first tasks run.
Neo shows you each task, its next runs and its destinations, then waits for your yes. You pick the schedule, Neo delivers to Slack or Gmail.
infrastructure
Every brand gets its own machine, running Claude Code.
Every brand has its own memory, accounts and data, with nothing mixed between clients. Neo reads, analyzes and proposes; it has no write access to your ad accounts.
Security and access →infrastructure
One machine per brand
We deploy a dedicated machine for each brand, with its own memory, data and tasks. Your team’s sessions, including those that cover several brands, run on the workspace machine.
model
Claude Code, on your subscription
Connect your Claude subscription and Neo runs on Claude Code with Anthropic’s most advanced models, including Claude Opus 5.5.
security
Your access stays yours
Creaboost never sees your Claude credentials, and your ad platform tokens stay in an isolated service.
go further
Every part of Neo has its own page.
questions
What teams ask us about Neo.
Which model does Neo run on?
Neo runs on Claude Code, on your team’s Claude subscription, with Anthropic’s most advanced models, including Claude Opus 5.5. Creaboost never sees your Claude credentials.
Does Neo change my campaigns?
Neo reads, analyzes and proposes. It has no write access to your ad accounts, and every decision stays yours: each platform’s review ends with 3 to 5 dated decisions for your team to apply.
Which sources does Neo read?
Every source you connect: Meta Ads, Google Ads, TikTok Ads, Microsoft Advertising, LinkedIn Ads, Snapchat Ads, Pinterest Ads and Reddit Ads; Shopify, GA4, Search Console, Merchant Center and Amazon; Klaviyo; and Instagram, Facebook and YouTube. Meta and Shopify are read live, the other sources are synced several times a day. Every metric is computed with the formulas of your Creaboost dashboards.
All integrations →How long does setup take?
A few minutes to connect your Claude subscription and your accounts. If you don’t have access, a link is enough to request it from whoever does. Neo then builds the brand memory, from the Brand Kit to the first concepts, without interrupting you: it saves its questions for the end and lets you know when the memory is ready.
Can I use Neo for several brands?
Yes. A workspace holds several brands, each with its own machine, memory and accounts. A chat, a board or a doc can cover a single brand, or several brands of the same workspace to compare them side by side.
Deploy Neo on your brand.
Book a demo and watch Neo work on your own accounts.