Meet Genie: AI Built for Real FileMaker Work
Genie in action inside the FileMaker workflow.
Genie started from a workflow many FileMaker developers already use: copy, paste, test, iterate.
The problem is that generic AI tools still struggle with FileMaker-specific structure, especially XML logic, schema dependencies, and script conventions. Results can be useful, but often need cleanup before they are safe to apply.
Genie is our answer: a FileMaker-aware shell with guardrails, review boundaries, and a practical apply workflow, delivered through a FileMaker plug-in.
What Genie is
Genie is not trying to be an AI model provider.
We’re building a FileMaker-first integration layer so the model you choose can work with better context, better validation, and more reliable output.
That means:
Better FileMaker-aware context handling
Predictable generation and validation flow
Clear review/apply boundaries before anything changes
What Genie does today
Understands context
Metadata-aware reasoning for scripts and structure-aware outputs
FileMaker-native conventions applied during generation
Applies safely
XML review and preview before apply
Preflight prompts when requests are incomplete
Two-stage pipeline for model control (reasoning/planning model first, generation model second).
Speeds up workflow
One-click apply back to FileMaker through the Genie plug-in
Copy, paste, and edit Script XML directly in the Script Workspace
Optional Web Mode for external reference lookup (shows source links)
Our goal
We’re building a fully integrated AI experience for real FileMaker development, not a demo layer.
Developers already use AI in other ecosystems because it improves speed and reduces repetitive work. We want that same productivity in FileMaker, without losing control or output quality.
We’ve already embedded core FileMaker rules, conventions, and structure into Genie. Our long-term vision expands beyond scripts into layouts, calculations, and broader solution workflows.
Roadmap
This roadmap is directional and may evolve as testing continues.
In active focus
Guided multi-step development flows (create/alter tables, create/debug scripts, multi-question runs)
Multiple script management
Multi-script paste/apply runs
Value List and Custom Function understanding
Chat history + streaming UI feedback
Reliability and quality
Regression tests for XML paste validity
Clipboard preservation via snapshot/swap
Improved function parsing accuracy
Better layout copy/paste handling for larger objects
Optional / exploratory
Output truncation / budget guard
Platform and model support
Platform support today: macOS (Windows planned)
Model-agnostic by design: OpenAI, Anthropic, Google, and local/custom setups