Oskar Architecten: Internal Knowledge Chatbot & Custom AI Agents
AI-powered internal assistant with RAG and custom agents for a leading Belgian architecture firm.
Case study
AI Knowledge Hub for Architecture
Oskar Architecten is a Belgian architecture firm known for education buildings, workspaces, and public projects. With a growing portfolio and expanding team, accessing the right information at the right time became a bottleneck. Project specs, regulatory guidelines, material references, and internal procedures were scattered across SharePoint, websites, and team knowledge.
We built an internal AI assistant that centralizes all this knowledge and makes it instantly searchable through natural conversation.
LibreChat Platform
The solution is built on LibreChat, providing a familiar chat interface that the team can access from any device. LibreChat’s multi-agent architecture lets us deploy specialized agents for different tasks, each with its own context, tools, and data sources.
RAG with Pinecone
At the core is a Retrieval-Augmented Generation (RAG) pipeline powered by Pinecone. Internal documents from SharePoint and content from reference websites are indexed and kept up to date. When a team member asks a question, the system retrieves the most relevant passages before generating an answer, ensuring responses are grounded in actual company data rather than generic knowledge.
Custom MCP Agents
We developed custom Model Context Protocol (MCP) servers that give the AI direct access to structured tools and data sources. These agents can search internal project databases, look up building regulations, retrieve material specifications, and more. Each agent is scoped to a specific domain, keeping responses focused and reliable.
Firecrawl for Web Intelligence
Using Firecrawl, the system continuously crawls and indexes relevant external websites, including supplier catalogs, regulatory portals, and industry reference sites. This keeps the knowledge base current without manual updates, and lets the team ask questions that span both internal and external sources.
The Impact
The assistant saves architects hours of searching through documents and websites. New team members get up to speed faster with instant access to institutional knowledge. And because the system cites its sources, the team can always verify and dig deeper into the original documents.