Capitalo - Multi-Market Financial Comparison Platform with AI-Powered MCP Agents

Showcase
1

Project Name

Capitalo - Multi-Market Financial Comparison Platform

Project Description

Capitalo is a production-ready financial comparison platform serving the DACH region (Germany, Austria, Switzerland) with domain-based localization (capitalo.de, capitalo.at, capitalo.ch).

We faced a massive challenge: migrating from a legacy WordPress/PHP platform with 500+ URLs, 100 banks, and 300+ financial products - while completely recreating all content with SEO and AI optimization. Timeline: 4 months.

The solution combines Directus as our headless CMS with a sophisticated MCP agent system powered by Claude Sonnet and Opus that automates content creation, SEO optimization, and quality assurance through 16 specialized AI agents.

Key Features

  • Multi-Market Architecture: Domain-based routing serving 3 markets (DE/AT/CH) with localized content

  • 96 Directus Collections: Polymorphic product model supporting 6 financial product types

  • 16 MCP Agents in 5 Tiers: From content creation to load testing, orchestrated by a tech-lead coordinator

  • Directus MCP Integration: AI agents manage content directly via Directus MCP

  • Directus Flows → n8n: Automated content pipeline triggered by Directus Flows

  • SEO + GEO Optimization: Content optimized for both search engines AND AI platforms (ChatGPT, Perplexity)

Challenge Entered

Project Challenge

Team Details

  • Alexander Senger - Founder & CEO of Capitalo

  • Niklas Senger - Project Support

Technical Details

Tech Stack

Layer Technology
CMS Directus 11+ with MySQL 8.0
Frontend Next.js 15.5, React 19, TypeScript 5.6
Styling Tailwind CSS 3.4, shadcn/ui
Hosting Directus on VPS (self-hosted), Next.js on Vercel
Automation n8n for workflow automation
AI Claude Sonnet & Opus via MCP

MCP Agent Architecture

We built 16 specialized AI agents organized in 5 tiers:

Tier 1: Orchestration

  • tech-lead-orchestrator - Coordinates complex multi-step tasks

Tier 2: Infrastructure

  • devops-deployment-agent - VPS, SSL, CI/CD

  • directus-setup-agent - Uses Directus MCP to create collections and Flows

  • data-migration-agent - WordPress † Directus migration

  • n8n-workflow-agent - Automation workflows

Tier 3: Frontend

  • nextjs-capitalo-agent- Next.js pages, ISR, calculators

  • frontend-developer - React components, Tailwind, shadcn/ui

Tier 4: SEO/GEO Pipeline

  • seo-content-agent- Creates content

  • geo-optimization-agent -Optimizes for AI platforms

  • seo-tech-agent - Implements structured data

Tier 5: Quality Assurance

  • design-review,playwright-tester, performance-optimizer, code-reviewer

MCP Servers Used

  • Directus MCP - Content management and Flow creation
  • Apify - Web scraping
  • Playwright - E2E testing
  • shadcn - Component installation

Content Pipeline Flow

  1. New bank added in Directus
  2. Directus Flow triggers n8n workflow
  3. n8n fetches data (website, Trustpilot, product DBs)
  4. Claude Sonnet creates SEO-optimized content via MCP agents
  5. GEO-optimization-agent restructures for AI platforms
  6. Result written back to Directus via Directus MCP
  7. Next.js frontend displays the content

Starting Point Details

We started with a fresh Directus 11 installation on a self-hosted VPS. Our legacy system was WordPress with a custom PHP backend and 71 MySQL tables.

Key Directus Features We Leveraged:

  1. Flexible Data Model - Polymorphic product schema handling 6 product types

  2. Translation System - Native multi-language support (de-DE, de-AT, de-CH)

  3. Directus MCP - AI agents create and manage content directly

  4. Flows & Webhooks - Trigger n8n workflows for automated content generation

  5. REST API - Perfect for Next.js ISR with automatic cache invalidation

  6. Self-Hosted - Full control, GDPR compliant on EU server

Migration Journey:

  • Weeks 1-2: Schema design and Directus setup

  • Weeks 3-4: Data migration scripts (71 tables → 96 collections)

  • Weeks 5-6: Frontend development with Next.js

  • Weeks 7-8: MCP agent system implementation

  • Ongoing: Content generation and optimization

Demo Video

Results

:white_check_mark: 80+ banks with AI-generated content
:white_check_mark: 300+ products (in progress)
:white_check_mark: 96 Directus collections
:white_check_mark: 16 agents
:white_check_mark: Directus MCP (VS Code with Claude Code + claude.ai)
:white_check_mark: 3 markets served (DE/AT/CH)
:white_check_mark: Trustpilot integration
:white_check_mark: SEO + GEO optimized content
:stopwatch: Completed in 3-4 months

Links

Built with :heart: using Directus, Next.js, and Claude AI

1 Like
2

Just a quick note to say thanks for submitting @Alexander_Senger! We’ll have winner’s announcements next week. Have a great weekend :sunflower: