AI agent operates FileMaker from plain English, narrating every click in a cloned voice
petrusenko_max · x · 2026-11-02
Developer petrusenkomax demoed an AI agent that takes a plain-English task — "open FileMaker, create a database, name it, save it" — and works out every click itself, with a visible cursor and narration in an AI clone of his own voice.
How it's built:
- Hermes Agent decides each action and verifies results before proceeding
- JSON action maps record positions of FileMaker's buttons, menus and dialogs
- Notion holds rules (popup handling, when to ask a human) and run logs
- Clicks are executed via macOS Accessibility (AppleScript); narration uses Kokoro TTS plus his cloned voice (F5-TTS) on a Mac mini, with word-level timing keeping speech and clicks in sync
- screencapture and ffmpeg merge screen recording and narration into one video, all on a separate Mac
A new task yields a new narrated video with no re-recording. Next steps: schema, value lists, and turning written procedures into training videos.
More from coding & agent
- Open-source AnyShot: a Claude agent acts as cinematographer, cuts 6-hour renders to 3 minutes — Scobleizer · 2026-11-02
- With Chrome riddled with UAF bugs, could an agent swarm finish Servo? — georgejrjrjr · 2026-11-02
- Dialkit macOS: open-source live inspector for tuning iOS UI from your Mac — mikelikesdesign · 2026-11-02
- Developer ships v0.6.83 entirely via autonomous agents, zero human input — jasonkneen · 2026-11-02
- OpenCode MAU jumps 28% to 22M in October, biggest growth since March — ycombinator · 2026-11-02
- Open-source pixel-plays: browser games as isometric pixel dioramas, one JS file, zero dependencies — algo_diver · 2026-11-02