Showcase · AI Agency

Seven agents, one working studio

What happens when you build a full agency out of AI agents? This: a pipeline that finds local businesses, analyzes them, builds a demo site and reaches out personally — with quality and cost monitoring at every step.

Already have a website? Then the same project starts at the Agentic Search Optimizer — no new site, just making your existing site discoverable for AI assistants.

7
specialized agents
9
industry demos live
23
tests on the deploy toolkit
24/7
pipeline keeps running
Architecture

Six specialists, one conductor

Every agent does one thing well and hands off a fixed, validatable format to the next. The orchestrator guards order, quality and budget.

Orchestrator playbooks · order · quality · budget Scout finds leads Diagnoser scans the site Builder builds the demo Filmer makes preview Pitcher reaches the lead Checker guards everything
finds

Scout

Finds local businesses via map and web data and validates leads against a fixed schema — including an agentic criterion.

analyzes

Diagnoser

Scans a business’s current site for tech, speed and AI-discoverability, and determines what can be improved.

builds

Builder

Generates a modern, personalized landing page in the industry style — ready to show.

shows

Filmer

Turns the new page into short video previews for the pitch.

reaches out

Pitcher

Writes and sends the personalized pitch, with the demo and video as proof.

guards

Checker

Checks quality and cost of every step; a cost monitor keeps the budget per lead in check.

Tech in detail

Discipline is what makes agents usable

One orchestrator above all

A single orchestrating agent directs the six specialists via fixed playbooks. Every step has its own prompt contract and output schema — no free improvisation in the pipeline.

Demo studio with 9 industries

design.neuralex.nl hosts nine client-ready industry demos with a personalization panel: company name, colors and photos are filled in per lead. The demo is the pitch.

Self-hosted LLM router

Every AI call runs through our own router, which picks the best model per task — any API pluggable, from EU-hosted to fully local — with cost and error monitoring per call.

Agent-ready toolkit

Reusable edge workers make every delivered site readable for AI assistants: llms.txt, schema.org, agent discovery. Covered by 23 automated tests, with recipes for the major platforms.

Lead validator & cost monitor

Every lead passes a validator with a fixed schema before any budget is spent. The cost monitor logs what a lead costs across the whole pipeline.

n8n as the backbone

The daily scout runs, retries and async steps run as n8n workflows alongside hundreds of other workflows in our lab — proven infrastructure, not greenfield.

Built with Multi-agent orchestrationSelf-hosted LLM routern8nEdge workersAgent-ready toolkitCost monitoring
Proof

Measured with our own yardstick

The first production site from this studio — ontwerp.neuralex.nl — run through our own Agentic Search Optimizer. Result on July 18, 2026:

100Discoverability 4/4
100Content 1/1
100Bot Access Control 2/2
100API, Auth, MCP & Skill Discovery 7/7
See it for yourself

The demos are real — click through them

Nine industry demos, built by the pipeline above. Every demo is a full, agent-ready site personalized per lead.