Home › Agents

Ultra SEO · AEO · GEOSystem inventory

54 agents. In 37 of them, the final call is made by code.

It is not one large model doing everything. They are specialised agents across seven families, and almost every agent that writes has a gate checking it afterwards, and that gate is almost always a pure function, not another prompt.

54

agents and gates

7

families

37

with a deterministic gate

0

publish without passing a gate

Where we measure

Ten engines running today

We would rather show what is live and what is coming than list ten logos and let the agency find out later which one returns no data.

Google

organic · every week

AI Overviews

AI summary · every week

ChatGPT

answer · every 15 days

Gemini

answer · every 15 days

Perplexity

answer · every 15 days

Bing organic

organic · every 15 days

Claude

answer · every 15 days

Copilot

answer · every 15 days

Grok

answer · every 15 days

Google AI Mode

AI summary · every week

The pattern that repeats

The AI proposes. The code decides.

What a model produces has to be verifiable by deterministic code, or it does not scale: generating a thousand pages with no gate is generating a thousand problems. That is why the AI almost never applies anything directly.

AIa model decides codedeterministic rule, no model AI + guardrailthe model proposes, the code validates and can refuse

Structure Critique

The AI suggests moving a term from pillar to satellite. The code validates every move and rejects anything that breaks the tree.

Reverse Probe

The AI answers what the brand is strong at. Extraction matches that answer against the project known vocabulary, instead of taking it on faith.

Integrity Gate

The AI writes. The code cuts the entire sentence when a number has no source. It does not soften it: it removes it.

The content line

From strategy to publication, with a gate at every step

No text goes live because the model thought it was good. It goes live because it passed four stages and gates that are code.

Context

What people ask, who already answers, which segment the company really operates in, and what the competition occupies.

Keyword Study
Sector Agent
Competitor Agent
Topics from the SERP

Briefing

Objective, question to answer, allowed sources and what cannot be said in that asset.

Story Brief
Pillar Organizer
Cluster Suggester

Writing

The text is born with the answer in the first sentence and the entity markup alongside, not after.

Blog Generation Engine
Page Pipeline
Hybrid Content

Gates

Six checks before publishing. One failure blocks the whole asset, and the reason shows on screen.

SEO gate for the text
Integrity Gate
Anti-Pattern Layer
Credential Guard

The catalogue

All 54, by family

Each one with the file path beside it. It was compiled by opening the repository, not from memory.

FAMILY 01

Measurement and monitoring on the AIs

7 agents

The engine that holds up everything else. Every number the agency shows its client comes from here.

Rank Tracker

Asks the same thing across the ten engines and records whether the brand appears, in what position, with which snippet, and which domains were cited instead.

server/rank-tracker.ts

AI + guardrail

Reverse Probe

It flips the question: instead of “does the client show up for this term?”, it asks the AIs “what is this client strong at?”. The result is matched against the project known vocabulary.

under construction

AI + guardrail

Sentiment Classifier

It reads the passage where the brand was cited and says whether the mention is positive, neutral or negative. It feeds the reputation screen.

server/sentiment.ts

AI Bot Radar

Records on the server when GPTBot, ClaudeBot or Perplexity read a page. A bot does not run JavaScript: the visit pixel would never see them.

server/ai-bot-tracker.ts

code

Authentic Bot Verifier

It verifies that the bot claiming to be GPTBot really came from OpenAI, against the published IP ranges. A fake bot would contaminate the metric.

server/traffic-origin/bot-verifier.ts

code

Cost Guard

No paid query happens without going through it: master key, accounting, client eligibility, monthly quota and the platform daily ceiling.

server/ai-search/cost-guard.ts

code

AI health

It catches a provider running out of credit or holding an invalid key before the functions start failing silently.

server/ai-health.ts

code

FAMILY 02

Audit and technical validation

7 agents

The part that answers “why is this site not cited”. It is the entry product.

Vynco AI Score

The site audit: 8 layers and 56 checks. It sweeps the homepage with two identities, one an ordinary browser and one GPTBot, to catch a firewall blocking the AI crawler.

server/score/engine.ts

AI + guardrail

SEO Quality Auditor

Runs 17 gates over the published pages and returns a 0 to 100 score with how many pages fail, examples and the recommended action.

server/seo-health-gates.ts

code

System Audit

Asks the database what matches, what does not, and what should be pulling from somewhere and is not. It runs on a click, it does not become a report that ages.

server/auditoria/checagens.ts

code

Scale Diagnostic

Turns a raw counter into a verdict, with what to do beside it. It is read mid-incident.

server/escala/diagnostico.ts

code

Agent-Ready

Publishes the endpoints that make every Vynco site speak the language of AI agents, following the well-known URI standards.

server/agent-ready.ts

code

Robots Policy

It decides whether the robots.txt being served allows or blocks crawling. A pure function on purpose: when this rule gets it wrong, it takes the client site out of Google.

server/lib/robots-policy.ts

code

Structure Critique

After the engine builds the pillar and satellite tree, a second pass assesses it like a strategist. The AI suggests moving a term, and the code validates every move.

under construction

AI + guardrail

FAMILY 03

Content quality gates

6 agents

No text reaches the client site without passing through here. Almost all of these are pure code.

SEO gate for the text

15 weighted criteria adding up to 100. This is the score that decides whether the text may be published.

shared/yoast-vynco.ts

code

Integrity Gate

It cuts the entire sentence when a number has no source. It does not soften it, it does not rewrite it: it removes it.

under construction

code

Anti-Pattern Layer

Catches model writing tics: em dashes, crutch phrases, generic openings and AI product names in client content.

server/lib/content-quality.ts

code

AI Reviewer Max

A second critical read of the final text, focused on depth and argument, not on keywords.

server/ia-max-reviewer.ts

Author Credential Guard

It prevents publishing content with invented authorship. The author has to exist and be verifiable outside the site.

server/author-credential-guard.ts

code

Compliance Guardian

It checks the creative asset against the brand restrictions before it leaves the studio.

server/creative-engine/compliance-guardian.ts

AI + guardrail

FAMILY 04

Strategy and research

10 agents

Before writing a single line: what people ask, who already answers, and what is missing.

Keyword Study Engine

It builds the business term tree by crossing commercial intent, market, and what the competition already holds.

server/lib/keyword-study.ts

AI + guardrail

Volume Estimator

It estimates demand by term and by neighbourhood where no reliable public data exists.

server/lib/keyword-volume.ts

Sector Agent

It reads the actual site content and works out which sectors the company really operates in, instead of taking the form at its word.

under construction

AI + guardrail

Competitor Agent

It finds out who is competing for the answer, not who the company thinks its competitors are.

under construction

AI + guardrail

Competitor Analysis

It opens the page that got cited instead of the client and says what it does differently.

under construction

Pillar Organizer

It splits terms between pillar and satellite by deterministic rule, with no model weighing in.

server/lib/organizar-pilares.ts

code

Cluster Suggester

It groups terms that answer the same question, so you do not generate five pages competing with each other.

server/ai-cluster-suggestor.ts

Topics from the SERP

It reads what is already published on the subject before proposing a story, so as not to repeat what exists.

under construction

AI + guardrail

Suggested Rules

Proposes business rules from the project facts; the code validates before they become configuration.

under construction

AI + guardrail

Term Translator

It turns technical jargon into the way a person actually asks.

server/lib/term-translator.ts

FAMILY 05

Writing and content creation

12 agents

Where the text is born. None of them publishes alone: they all feed into family 03.

Page Generation Pipeline

It builds the page that answers an unowned question, with the entity markup already attached.

server/page-generation-pipeline.ts

AI + guardrail

Blog Generation Engine

It writes the article from the brief, with the structure the machine reads and the answer in the first sentence.

server/blog/generation-engine.ts

AI + guardrail

Story Brief

Turns the topic into a brief: objective, question to answer, sources and what cannot be said.

under construction

AI + guardrail

Hybrid Content

It combines the generated text with an excerpt from a verifiable source, marking which part came from where.

server/lib/hybrid-content.ts

AI + guardrail

Matrix Enricher

Fills the content matrix by region and service, pointing out what is left to cover.

server/aio-matrix-enricher.ts

Living Content

It refreshes data that has aged inside an already published page. It never rewrites the whole text.

under construction

AI + guardrail

About, rewritten

Proposes a version of the about page without replacing the original, for side-by-side comparison.

under construction

Topic Generator

Suggests the next subject from the measured gap, not from a generic editorial calendar.

server/ai-blog-topic.ts

Instruction Editor

You write what you want changed in plain language and the code applies it to the structure, without breaking the markup.

server/lp-ai-editor.ts

AI + guardrail

Catalogue Curator

It classifies and describes store products at scale, with a coherent category.

server/ecommerce-curator.ts

Blog Autopilot

It keeps the blog line running on its own within the configured limits.

server/ecommerce-blog-autopilot.ts

Review Responder

Drafts replies to map reviews, which are a citable source for the model.

server/jobs/google-reviews-poll.ts

FAMILY 06

Creative studio

9 agents

The visual asset for the ad and for social, on the same pattern: the AI proposes, the orchestrator decides.

Intent Extractor

It works out what the asset needs to say before anyone thinks about an image.

server/creative-engine/intent-extractor.ts

AI + guardrail

Clarifying Dialogue

Asks for what is missing instead of inventing what was not said.

server/creative-engine/clarification-dialog.ts

Reference Analyser

It reads the reference the agency sent and extracts the pattern, not the copy.

server/creative-engine/reference-analyzer.ts

Concept from Reference

Proposes the asset concept based on the extracted pattern.

server/creative-engine/concept-from-reference.ts

Creative Director

It picks the direction and rejects anything that strays from the brand.

server/creative-engine/director-claude-opus.ts

AI + guardrail

Layout Designer

It lays out the design inside the grid, with hierarchy and spacing set by rule.

server/creative-engine/claude-designer.ts

AI + guardrail

Carousel Generator

Breaks the argument into cards, one per idea.

server/creative-engine/carousel-generator.ts

Image Router

It decides which image engine to use for each type of asset.

server/creative-engine/image-router.ts

Pipeline Orchestrator

It drives the sequence, holds the state and blocks any asset that has not been through every step.

server/creative-engine/pipeline-orchestrator.ts

code

FAMILY 07

Social and continuous operation

3 agents

What keeps the pipeline running with nobody pressing a button.

Social Planner

It builds the calendar from the subject the measurement raised, in three formats.

server/social/social-planner.ts

AI + guardrail

Autopilot Processor

It runs the day queue respecting the limit, the order and the window.

server/jobs/blog-autopilot-processor.ts

code

Automation Gate per Project

It decides whether that project may run automatically right now. When in doubt, it does not run.

server/jobs/project-automation-gate.ts

code

Start by measuring. It is free.

In minutes you see which engines cite your brand, which do not, and who appears in your place.

No card, no commitment.