The problem: information overload and the barrier of complex interfaces
Today's digital resources — whether large municipal portals, sizeable corporate knowledge bases or online booking platforms — hold colossal volumes of information. To find what they're looking for, users have to navigate multi-level menus and complex filtering systems. For many visitors — especially older generations, or anyone browsing your site from a smartphone, on the go — having to manually set 5 to 7 filter parameters becomes an insurmountable barrier, costing your platform its audience.
Kyzym.fr's solution: an AI assistant based on the Model Context Protocol (MCP)
Specialised in complex web development, I offer the integration of an intelligent agent that changes the interaction paradigm with your digital products. Rather than learning an interface, we let the user simply type or say, in natural language, what they want to find.
The assistant connects to your infrastructure via a purpose-built MCP (Model Context Protocol) server. It's not a simple "external chatbot" — it's a deeply integrated module that understands the structure of your databases and extracts reliable, one-hundred-percent accurate and up-to-date information from them.
Universal integration: from CMS to microservices
This architecture isn't limited to classic monolithic websites. The technology is flexible and scalable :
Classic CMS and monoliths. The assistant integrates easily with robust content management systems like TYPO3, with Symfony and PHP projects, and other popular platforms, connecting directly to their databases.
Microservices architecture. The full power of MCP shows in complex distributed systems. The AI assistant becomes a single entry point (hub) for multiple independent microservices. For example, to handle a single user request, the system can query the events microservice, the ticket booking service and an external CRM in parallel. To the user, it all feels like one smooth, natural conversation, while behind the scenes the AI consolidates data from multiple sources.
How it works technically
The architecture adapts to your goals, budget and strict security requirements — including European GDPR standards.
Cloud API solutions. Integration with the market's most advanced models (Gemini, OpenAI) for a fast launch and maximum performance.
Local LLMs on dedicated VPS. For projects working with confidential documents, I deploy locally-hosted AI models on secure servers. All data stays strictly within your infrastructure, with no transmission to third-party APIs.
Application scenarios: practical use cases
Case 1 — the smart listing board of a municipal portal
The traditional path: the user has to find the date filter, select "15 to 18 October", open the category dropdown, tick "Exhibitions" and locate a specific speaker.
The path via the AI assistant: the user simply types: "I'll be in town from the 15th to the 18th, I want to see exhibitions by [Name]".
Result: the AI analyses the request, the MCP server runs a precise query against the events microservice, and the user instantly receives a ready-to-use selection, with direct links to ticketing.
Case 2 — navigating corporate and municipal documents
Instead of a tedious search through multi-level archives, the user simply asks: "Where can I download the micro-business registration form for 2024?". The assistant instantly provides the right PDF file along with a short usage guide, drawing exclusively on your project's internal database.
Key benefits for your business
Radical improvement in user experience (UX): reduced bounce rate. Your visitors get a useful answer immediately.
Inclusivity and accessibility: an ideal solution for people who struggle with small interface elements, complex forms or multi-select menus.
Accuracy, with no hallucination: thanks to the MCP protocol, the artificial intelligence relies exclusively on the real data in your databases and microservices, never inventing information that doesn't exist.
Ready to equip your website?
Tell us about your business — we'll show you what an AI assistant can do for you. Contact us.