Last updated: 2026-08-19
Skilly is a screen-watching AI voice tutor for macOS that teaches 6 creative and developer apps, including Blender, Figma, and After Effects, through spoken guidance. It streams microphone audio and live screenshots to OpenAI's Realtime API over a single WebSocket call, then speaks back while moving an on-screen cursor to the exact UI element it is describing.
About Skilly
Skilly is a voice-first AI tutor for macOS built by indie developer Mohamed, released as an open-source fork of Farza Majeed's Clicky project under the Apache-2.0 license. The app lives in the Mac menu bar and pairs a live screen feed with OpenAI's Realtime API, so the tutor can see what you're doing and answer out loud without you ever switching windows.
The core mechanic is a pluggable teaching layer: every supported app gets its own SKILL.md curriculum, so the coaching stays specific to the software in front of you instead of offering generic tips that don't match what's on your screen.
Skilly for Builders extends the same voice-and-cursor mechanic to a company's own product: product owners load their own documentation into a Studio project, lock the widget to their domain, and ship a branded onboarding guide inside their web app without exposing provider credentials to the browser.
Pricing
Personal plan: free 15-minute trial (no card), then $19/month for 3 hours of tutoring. Builders plan: free tier at 30 minutes/month, Starter $29/month (400 minutes), Studio $99/month (1,500 minutes), Scale $299/month (5,000 minutes). Sessions pause once the monthly minute cap is reached.
| Tier | Monthly price | What it includes |
|---|---|---|
| Free Trial | Free | |
| Personal | $19/mo | |
| Builders Starter | $29/mo | |
| Builders Studio | $99/mo | |
| Builders Scale | $299/mo |
Key Features
- Voice-First Screen-Aware Tutoring: Streams your microphone audio and live screenshots to OpenAI Realtime in a single WebSocket call, cutting AI response latency roughly in half compared to traditional three-stack TTS+STT+LLM setups.
- Animated Cursor Pointer: Calls a point_at_element tool to physically animate a blue cursor to the exact button, panel, or control mentioned in the explanation, working across all connected monitors without requiring the user to identify it manually.
- Per-App Teaching Curricula: Ships with SKILL.md curriculum files for 6 apps (Blender, Figma, After Effects, Xcode, Photoshop, VS Code), each containing staged lessons, UI vocabulary, and completion signals so teaching is domain-specific rather than generic.
- Always-On Live Tutor Mode: Captures screenshots only while the user is speaking, so the Mac screen is not continuously recorded; users work normally and speak when they need guidance, with push-to-talk also available as an alternative.
- Progress Tracking Across Sessions: Saves completed lessons and curriculum stages per app locally so each new session resumes exactly where the previous one left off, rather than restarting from the beginning.
- Auto-Language Detection: Detects the spoken language in real time (English, Arabic, Japanese, Chinese, Spanish, and more) and replies in the same language with no settings to configure.
- Skilly for Builders Embeddable Widget: Lets SaaS and website owners train Skilly on their own docs via a Studio project, then deploy a branded voice-and-pointer onboarding guide on their domain, with a paid plan for teams that outgrow the free tier.
- Fully Open Source (Apache-2.0): The complete codebase is public at github.com/tryskilly/skilly, allowing teams to self-host with their own OpenAI API key and inspect the code before putting it in front of users.
Pros
- The free trial requires no credit card at all, a lower barrier to entry than most competing AI subscriptions that make you pay before you've tried anything.
- The latency win over a traditional three-stack pipeline is genuinely noticeable in practice: replies feel closer to a live conversation than the stop-start pacing of most voice assistants.
- Animated cursor that physically points to UI elements on-screen is a differentiated capability no browser-tab AI assistant (ChatGPT Voice, Gemini Live) can replicate for native Mac apps.
- The permissive open-source license lets security-conscious teams audit the whole system, or self-host it entirely on their own infrastructure with their own OpenAI key, before ever putting it in front of end users.
Cons
- Personal desktop tutor is macOS-only; Windows and Linux users have no equivalent product in 2026.
- Monthly tutoring time is capped for the personal plan, and power users who want daily practice sessions have no overflow option once they hit the limit.
- Relies entirely on OpenAI Realtime API, so any OpenAI outage or latency spike directly affects Skilly's reliability with no fallback provider option documented.
- Teaching curricula exist for only 6 apps; learning Adobe Premiere Pro, DaVinci Resolve, or any non-listed tool requires writing a custom SKILL.md file.
Data Handling
- Training-data policy
- OpenAI Realtime API processes audio and screenshots during sessions. Skilly's own servers do not train on user session data. Provider API keys remain on Skilly's backend infrastructure rather than being exposed in the browser or SDK.
Frequently Asked Questions
What are Skilly's pricing plans in 2026?
The Personal plan runs $19/month for 3 hours of tutoring, with a short trial before billing starts. Skilly for Builders scales separately: Starter is $29/month for 400 minutes, Studio is $99/month for 1,500 minutes, and Scale is $299/month for 5,000 minutes, with minutes pooled per plan and sessions pausing once the monthly cap is hit. None of the Builders tiers charge per seat.
Can you use Skilly without paying?
Skilly's personal macOS tutor gives you a 15-minute trial, and Skilly for Builders currently offers a genuinely free tier of 30 minutes a month during its early-access period. Past those free minutes, a paid plan is the only way to keep tutoring sessions going. The codebase is also open source, so self-hosting it with your own OpenAI API key is free of Skilly's own charges, though OpenAI's usage fees still apply.
Which tools compete with Skilly in 2026?
Gemini Live is the closest cross-platform alternative if you don't need an app-specific cursor pointing at UI elements. Clicky, the MIT-licensed project Skilly was forked from, is worth trying if you want the simpler original push-to-talk flow without curriculum tracking. For coding-specific help inside an IDE rather than a general creative-app tutor, Cursor is the stronger pick.
Skilly or ChatGPT Voice: which should you pick?
ChatGPT Voice runs on macOS, Windows, iOS, and Android and costs $20/month for its unlimited Pro plan, but it cannot see your screen or point at anything on it. Skilly only runs on macOS and caps personal tutoring at a few hours a month, trading platform reach for screen awareness: it watches what you're doing and physically moves a cursor to the exact button or menu it's talking about. Pick ChatGPT Voice for general-purpose conversation on any device; pick Skilly when you need hands-on guidance inside a specific Mac app.
How long does it take to get going with Skilly?
A first Skilly session starts within a couple of minutes of downloading it: grab the macOS app from tryskilly.app and open it from the menu bar to begin the free trial. Pick a supported app such as Blender or Figma, start talking, and Skilly listens, captures your screen, and speaks back while pointing its cursor at what it means. Teams building Skilly for Builders instead create a free Studio account, upload their own docs or a website URL, and copy a JavaScript snippet onto their site.
Top Alternatives
- Cursor: Cursor puts AI coding help inside an IDE for people already writing software. Skilly instead watches the whole screen and talks you through design or developer tools you're still learning, step by step.
- ChatGPT: ChatGPT Voice runs everywhere at $20/month but has no idea what's on your screen. Skilly stays Mac-only and limits how many tutoring hours a personal plan includes each month, trading that reach for a cursor that physically points at the UI element it's explaining.
- Claude: Claude is a capable text-based assistant that runs in any browser, on any OS. Skilly stays Mac-only but adds something Claude doesn't offer at all: real-time voice tutoring that sees the screen and points at it.
HokAI guides covering Skilly
- What Skilly Is Actually For, and Which Skilly You Mean: Skilly is filed as an AI voice assistant next to Siri, but it is really two small products forked from Clicky in five days. Here is which one is worth using.
- Murf AI vs Skilly: Why This Comparison Is Asking the Wrong Question: Murf AI makes voiceover audio and voice agents; Skilly is a $19/month Mac tutor that narrates Figma and Blender live. Here's which job needs which tool.