Building Your Own Work Integration Hub with GitHub Copilot CLI

GitHub's Staff Software Engineer Brittany Ellich built a personal work integration hub that reduces digital fragmentation by bringing together information scattered across multiple apps into one place. The goal was to unify scattered tools into a single calm central screen, reducing the burden of context switching between thoughts and tasks.

The core approach is plan-then-implement. First, Copilot asks questions to thoroughly organize how things should work and what the requirements are, and then Copilot takes on implementation based on that plan. Thanks to this flow, v1 was completed to a workable level in just one day, even while handling other work in parallel.

The development approach is also clear. Brittany works synchronously in VS Code's agent mode, running up to 2 non-conflicting agent workflows in parallel. At the same time, she offloads asynchronous tasks like bug fixes or clearly-scoped tech debt to Copilot Cloud Agent, while keeping direct control locally over work that needs hands-on verification.

She wasn't particularly attached to specific technology choices. She had originally wanted to try building an Electron app, and it did become her first Electron project, but she says Agent Mode implemented most of it. Instead, she put in significant effort simplifying the repository for public release, and she also came to feel that agents are good at adding code but less proactive about reducing it.

To reproduce this, you'll need the following combination:

• Electron: desktop app framework
• React: UI and state management
• Vite: fast build tool
• Tailwind: styling
• WorkIQ MCP: MCP server and CLI for accessing Microsoft 365 data

For the runtime environment, you'll need Node.js 18+, GitHub Copilot CLI (for WorkIQ setup), a Microsoft 365 account (for calendar sync), and an ElevenLabs account (for voice assistant setup). Brittany concludes that a personal project aimed at solving a small annoyance ends up growing into the most useful system, and is also the best starting point for learning a new AI tool.