Honest fork story
Skilly vs Clicky
Skilly is an open-source fork of Farza Majeed's Clicky. Same macOS cursor-buddy DNA. We added a live tutor loop, a single-call Realtime pipeline, a pluggable skills layer, and multi-language support on top. This page is the full diff so you can decide which one fits your workflow.
Credit where it is due. Clicky is the base. Farza shipped it open source (MIT) and it is the reason this whole category exists. Skilly would not be possible without that work. If the additions below are not useful to you, Clicky is the right tool — go use it. Clicky is MIT, Skilly is Apache-2.0, both on GitHub.
What Skilly adds on top of Clicky
Live Tutor mode
Clicky is push-to-talk only. Skilly adds an always-listening Live Tutor mode where the screen is captured only during voice activity — so you can keep working and just speak when you need help.
Single-call OpenAI Realtime API
Clicky stacks three providers: TTS + STT + LLM. Skilly swaps that for a single direct voice-to-voice call against OpenAI Realtime. Roughly halves response latency in practice and simplifies the key surface.
Pluggable Skills layer
Markdown-based per-app curriculum (Blender, Figma, Xcode, Photoshop, After Effects, VS Code) that lives outside the model weights. Stays current even when the underlying model training cutoff is stale.
Course-style progress tracking
Tracks which skills / lessons you have completed per app, so tutoring picks up where you left off instead of starting from scratch every session.
20+ language auto-detect
Realtime model picks the language you spoke in and responds in the same language — no config, no menu. Works across English, Arabic, Japanese, Chinese, Spanish, and more.
You want the tutor experience — per-app curriculum, progress tracking, lower-latency voice, and multi-language out of the box. You would rather install one app than configure a three-stack voice pipeline yourself.
You want to control the entire pipeline — bring your own STT/LLM/TTS providers, tweak prompts, read every line of code. Clicky is a cleaner base if you plan to fork it yourself rather than use it as-is.
Feature by feature
| Feature | Skilly | Clicky |
|---|---|---|
| Voice pipeline | Single-call OpenAI Realtime API (voice-to-voice) | Three-stack: STT → LLM → TTS (cloud providers you configure) |
| Always-listening mode | Yes — Live Tutor mode (voice-activity gated) | No — push-to-talk only |
| Per-app curriculum | Yes — Skills layer with Markdown SKILL.md per app | No — prompt defines capability, same for every app |
| Progress tracking | Yes — course-style progress per skill | No — stateless |
| Languages | 20+ auto-detected via Realtime model | Depends on chosen STT/TTS providers, manual config |
| Cursor guidance | Yes — cursor flies to the exact UI element | Yes — core Clicky behavior, inherited |
| Platform | macOS 14.0+ | macOS (upstream repo) |
| License | Apache-2.0 | MIT (upstream) |
| Pricing | 15 minutes free, $19/month for 3 hours tutoring | Free, open source — bring your own API keys |
Common questions
Is Skilly a Farza fork?
Yes. Skilly is forked from Farza Majeed's clicky (github.com/farzaa/clicky, MIT, 5,332 stars) and credits clicky in the README + LICENSE chain. The fork is at github.com/tryskilly/skilly under Apache-2.0. Skilly was built in 5 days on top of clicky's macOS plumbing in early 2026.
What did Skilly add on top of clicky?
Four things: (1) Single-call OpenAI Realtime pipeline replacing clicky's TTS+STT+LLM 3-stack — sub-second latency vs ~3 sec. (2) Pluggable Skills layer — Markdown curriculum per app (Blender, Figma, AE, DaVinci, Premiere — ~33 hours total, free). (3) Course-style progress tracking with auto-advance signals. (4) 20+ language auto-detect via Realtime model — no settings menu, no model picker.
Does Skilly compete with clicky?
Not really. Clicky is a more general-purpose, hackable foundation aimed at developers who want to fork it themselves. Skilly is opinionated for the "voice tutor for unfamiliar apps" use case. Both projects are open source and credit each other; the relationship is sibling/parent, not competitive. If you're a builder who wants to ship your own voice + screen tool, fork clicky directly. If you want a finished tutor app, install Skilly.
Can I run Skilly without paying $19/month?
Yes — Apache-2.0 source means you can clone github.com/tryskilly/skilly, build locally, and run with your own OpenAI API key (BYOK option, in beta). You'd pay OpenAI directly per minute of Realtime usage instead of the $19/month flat. The hosted $19/month tier exists for convenience: no API key management, no per-minute billing surprises, founder-beta price lock for life on the first 50 seats.