
Desktop AI buddy for Mac and Windows that sees your screen on a hotkey, talks you through any app and runs agents you start by voice. Y Combinator Spring 2026. Free plan, plus Pro and Max.
HeyClicky suits Mac and Windows users who would rather be shown than search: press a hotkey, ask out loud, and it answers in speech and points at the button. Since October it can also drive your mouse with permission and work inside Chrome. Agent-heavy users end up on Max, and there is no Linux version.
Visual cursor pointing on any Mac app with real-time screen context, eliminating context-switching entirely for voice-driven assistance.
A fast release cadence and a consumer-first voice interface; the app is now closed source, and Apple or Microsoft could build the same screen help into the operating system.
Main trade-off
Tight agent message allowances, no SOC 2, and a Windows app only weeks old.
| Plan | Monthly price | What it includes |
|---|---|---|
| Free | Free | No card needed; 25 talk and 25 agent messages a month, plus dictation |
| Pro | $20/mo | Yearly option at $16 a month; talk and dictation without limits, 150 agent jobs per month |
| Max | $100/mo | Yearly option at $80 a month; talk and dictation without limits, 1,000 agent jobs per month |
Three plans. Free: $0, no card, 25 talk messages and 25 agent messages a month plus what the vendor calls plenty of dictation. Pro: $20 a month, unlimited talk and dictation, 150 agent messages a month. Max: $100 a month, unlimited talk and dictation, 1,000 agent messages a month. Yearly billing takes 20% off, which works out at $16 and $80 a month. Students get 50% off their first month of Pro, new or existing. Everything is month-to-month and cancels in one click with access to the end of the cycle. When agent messages run out, talk keeps working and agents resume on the next cycle. Read on heyclicky.com on 11 November 2026.
| Feature | Free | Pro | Max |
|---|---|---|---|
| Price, monthly billing | $0 | $20/month | $100/month |
| Price, yearly billing | $0 | $16/month | $80/month |
| Talk (spoken conversation) messages | 25/month | Unlimited | Unlimited |
| Dictation | Included | Unlimited | Unlimited |
| Agent messages | 25/month | 150/month | 1,000/month |
| Card required to start | No | Yes | Yes |
HeyClicky is a desktop assistant for macOS and Windows from Humansongs, Inc., a Y Combinator Spring 2026 company. Press a hotkey and it captures your screen, then speaks step-by-step guidance while pointing at what to click. It talks in more than 130 languages, and every request is processed in the cloud by OpenAI, Anthropic or Google models.
Screen capture on a hotkey
A push-to-talk shortcut grabs the screen at the moment you ask, and the vendor says it never looks at your screen at any other time.
Spoken guidance in 130+ languages
Answers come back as speech in more than 130 languages, with 30 regional voices added on 30 October 2026, so your eyes stay on the app you are using.
Drawing on your screen
Since version 1.1.1 in October 2026 it circles, points and sketches in 2D, and builds 3D models you can rotate when the answer is a physical object.
HeyClicky drive
With your yes it moves the mouse and types in an app for you; moving the mouse, talking or pressing Escape hands control back at once.
Background agents
Saying the agent phrase starts a job with its own window; since 30 October a long job keeps going instead of pausing and still counts as one agent message.
Chrome extension for agents
From the Chrome Web Store, it lets agents work in your signed-in tabs in Chrome, Edge, Brave, Arc or Vivaldi, and it never fills password or card fields.
Routines by voice
Saying something like "every Tuesday and Thursday at 5, check my inbox for invoices" sets up a repeating job, and since 6 November an existing job can be made to repeat from its own chat.
Custom MCP connections
Since version 1.0.49 in September 2026 you can add your own MCP servers by URL or local command, next to the built-in app catalogue and several accounts per app.
Mac and Windows on one account
The Mac app needs macOS Sonoma 14.2 or later and the Windows app runs on Windows 10 or later, with one account and one plan covering both.
Most users expect these and they are not currently supported.
| Tool | Price from | Free plan | Runs on | Rating |
|---|---|---|---|---|
| $20/mo | Yes | Mac · Windows | 4.1 | |
| $8/mo | Yes | Web · iOS · Mac · Windows | 4.4 | |
| $15/mo | Yes | iOS · Android · Mac · Windows | 3.9 | |
| $20/mo | No | Web · iOS · Android · Mac · Windows | 4.6 |
Proprietary app under HeyClicky terms. The original Clicky prototype at github.com/farzaa/clicky stays MIT-licensed, but new work has been private since 27 April 2026.
There are three plans. Free costs nothing and needs no card. Pro is $20 a month with unlimited talk and 150 agent messages, and Max is $100 a month with 1,000 agent messages. Yearly billing takes 20% off, and students get half off their first Pro month.
Yes, you can use it without paying. The Free plan includes dictation and a monthly allowance of 25 talk and 25 agent messages on Mac and Windows, with no card. Spoken questions never stop; only agents pause until the next cycle.
It only captures your screen when you press the hotkey, does not store screenshots on its servers and says neither it nor its AI providers train on your data. Requests are processed in the cloud in the US, conversations are kept until you delete the account, and it has not yet completed a SOC 2 or ISO 27001 audit.
The app you download is not. Farza's farzaa/clicky repository is MIT-licensed, but it holds the original Clicky prototype, and since 27 April 2026 new work is private. You can build that prototype in Xcode with your own API keys; official downloads come only from heyclicky.com.
Y Combinator, through its Spring 2026 batch; YC's standard deal puts $500,000 into every company it accepts. The company names no further investors, and the funding database VCBacked records an undisclosed June 2026 pre-seed round backed only by YC.
HeyClicky is a desktop assistant for Mac and Windows that looks at your screen only when you press its push-to-talk hotkey. Ask out loud and it works out what you are looking at, answers in speech, and draws on the screen or moves a pointer to the control you need. Say "heyclicky agent" and the same request becomes a background job that can use the apps you connect. Farza Majeed started it in 2026 as Clicky, an open-source Mac prototype, and took the company, Humansongs, Inc., through Y Combinator's Spring 2026 batch. The homepage claimed more than 25,000 users when we read it in November 2026.
The GitHub question has a two-part answer. The repository at github.com/farzaa/clicky is real and carries the MIT licence, but its README says that since 27 April 2026 all new work stays private. On 11 November 2026 we checked it: 7,987 stars and 1,597 forks, no code change after 27 April, and one later commit that only rewrites the README to point at the official downloads. What you download from heyclicky.com is the official app, and its code is not public. The repository holds the original prototype, which you build yourself in Xcode with your own API keys and a Cloudflare Worker, and the README warns that Windows apps built from it are community ports.
On funding, Y Combinator puts a standard $500,000 into every company it accepts, split into $125,000 for 7% and $375,000 on an uncapped SAFE. HeyClicky has announced no other round, and VCBacked lists an undisclosed pre-seed round dated June 2026 with Y Combinator as the only investor.
October changed what the product does. Version 1.1.1 on 23 October brought the first Windows build, drawings in 2D and 3D, and HeyClicky drive, which moves the mouse and types only after you say yes and hands control back the moment you press Escape. A Chrome extension, now in the Chrome Web Store and also working in Edge, Brave, Arc and Vivaldi, lets agents work in tabs where you are already signed in. Routines can be set up by voice, and version 1.1.3 on 6 November lets agents upload files to websites after asking once. Custom MCP connections sit next to a built-in app catalogue; there is still no public API.
Nothing is processed on your machine. Requests go to models from OpenAI, Anthropic and Google, with Cerebras, Groq and Deepgram handling dictation, and the privacy policy updated on 10 November says your data is processed in the United States. Screenshots are not stored on HeyClicky's servers, but conversations and a memory of what it learns about you are kept until you delete the account. Hokai files it with agentic AI tools and Mac apps.
Who should look elsewhere: anyone whose screen content must stay on the device, any team that needs a SOC 2 report, and Linux users. ChatGPT reads code and notes from a fixed list of Mac apps without taking screenshots, Microsoft Copilot gives spoken screen help on Windows with a Microsoft 365 plan, Raycast adds screen context to a launcher, and Wispr Flow covers dictation on its own. The HeyClicky alternatives guide sorts them by job. Everyone else, above all someone learning an unfamiliar app on a personal computer, should start on the free plan and pay for Pro only once the free talk allowance runs out.
Product facts on this page (pricing, plans, integrations, compliance) are read from HeyClicky's own pages and documentation. The verdict is Hokai's reading of them, not the vendor's claim.