Presenters
Source
From Quarters to Minutes: Building a Production-Ready IDP with AI 🚀
How long does it take to build a Minimum Viable Product (MVP) for an Internal Developer Portal (IDP)? If you ask most platform engineers, you will hear a range of answers: one month, six months, or even a full year.
Matan Peles, Field CTO at Port.io, has seen this struggle firsthand while working with giants like GitHub, PwC, and CyberArk. He argues that if you do not approach the problem correctly, building a portal can consume quarters of your roadmap.
But what if you could compress a year of work into just five minutes? Using an AI-first platform and the Model Context Protocol (MCP), Matan demonstrates how to build a fully functional IDP in the time it takes to brew a cup of coffee. ☕
🏗️ The Architecture of a “Coffee-Powered” Company
To prove the power of automation, Matan uses a hypothetical global coffee distribution company as his test case. The scale is non-trivial:
- 200 engineers 👨💻
- 40 microservices ⚙️
- Global expansion goals 🌐
- A tech stack involving GitHub, Kubernetes, and AI agents 🤖
The challenge? Information is scattered everywhere. Developers need a single pane of glass to find documentation, Slack channels, and service owners without hunting through dozens of repositories.
🛠️ The Magic Behind the Curtain: Port and MCP
The “magic trick” relies on Port, an AI-first developer portal platform. Matan uses the Model Context Protocol (MCP) and APIs to translate a simple text prompt into a complex data architecture.
1. Defining the Data Model (Blueprints) 📋
Matan starts with an empty environment—the nothing up my sleeves moment. By feeding a descriptive prompt into Port’s MCP, the system automatically understands the company’s architecture. It immediately begins creating Blueprints, which are digital representations of resource types like:
- GitHub Repositories 💾
- Kubernetes Clusters ☸️
- Microservices 🛠️
- AI Agents and Skills 🤖
2. Connecting the Ecosystem 🔌
Once the data model exists, the AI uses tokens to connect directly to GitHub and Kubernetes. It does not just create empty boxes; it syncs and discovers real information, bringing live metadata into the portal.
🚀 The Result: A Fully Functional Developer Hub
In less than five minutes, the empty environment transforms into a high-utility dashboard. The portal now features:
- A Comprehensive Service Catalog: This displays all services live from GitHub, complete with metadata, Slack channel links, and documentation.
- Infrastructure Visibility: The portal shows connected Kubernetes clusters and discovered repositories.
- AI Registry: A unique section for AI agents, skills (extracted from MD files), and MCP servers, providing the team with total visibility into their AI tools.
- Self-Service Marketplace: Instead of opening tickets, developers can now:
- Create a new microservice (including CI/CD and repo setup) with one click. 🖱️
- Deploy applications to specific target environments.
- Provision cloud resources on demand.
💡 Interacting with Your Infrastructure
The portal is not just a static display; it is an interactive partner. Matan demonstrates the Command + I side chat, which allows developers to:
- Ask complex questions about the envisioned architecture. ❓
- Run actions directly through the chat interface.
- Consume the portal’s data through the same MCP that built it.
🎯 The Bottom Line
The traditional tradeoff for an IDP has always been time vs. customization. Matan Peles proves that by leveraging AI-first platforms and MCP, you can bypass the “year-long MVP” trap.
You no longer have to choose between speed and depth. You can build a structured, automated, and self-service-ready portal that scales with your 200-engineer team—all before your Iced Americano gets warm. 🧊☕
“If you don’t do it right, getting to an MVP can take quarters if not years. We just did it in five minutes.” — Matan Peles