OmniWrite – Your System-Wide Flutter Butler

The "Butler" Stack: Using Serverpod 3 to orchestrate Gemini AI and Flutter with native macOS bridges for a fast, invisible assistant.
OmniWrite in action: A seamless Flutter-built overlay that provides instant, system-wide AI text transformations without leaving your app.

Inspiration

Writing today happens everywhere — emails, documents, code editors, browsers — but improving text usually means switching apps, copy-pasting, and breaking focus. I wanted a system-wide assistant that works wherever you write, instantly.

That idea became OmniWrite: a one-click Flutter Butler that transforms text in place.

What it does

OmniWrite lets users select any text on their system and instantly:

Fix grammar
Rewrite content
Change tone (formal, friendly, concise, etc.)
Translate between languages

All without leaving the current app.

How I built it

OmniWrite is built as a Flutter-first application with deep OS integration:

Flutter provides the main UI and user experience.
A native macOS Swift overlay enables system-wide text selection and quick actions.
Serverpod acts as the backend orchestration layer:
- Receives transformation requests from Flutter
- Handles prompt construction, validation, and processing
- Integrates with the Gemini API for AI-powered text transformations
Flutter communicates with the native layer using method channels, keeping the app fast and lightweight.

This architecture allows the frontend to stay simple while Serverpod centrally manages all AI behavior.

Challenges

Designing a system-wide experience that feels instant and non-intrusive
Bridging Flutter with native macOS APIs reliably
Keeping AI processing consistent and controllable through the backend
Ensuring Serverpod remained a core part of the system rather than just a proxy

What I learned

How powerful Flutter can be when combined with native platform features
How Serverpod simplifies building a clean, Dart-based backend for real-world apps
The importance of backend orchestration when integrating AI into user-facing tools