HeyClicky is a Mac-native AI assistant that sits next to the cursor, watches the screen, and listens for voice commands. This launch adds two dictation modes aimed at people who write across a lot of different apps, from Claude Code and terminals to Gmail, Slack, LinkedIn, and Google Slides. Founded in 2026 by Farza Majeed, who previously built Buildspace into a community of 125,000+ builders , the company is part of Y Combinator's Spring 2026 batch.
The first mode is a plain speech-to-text that streams into whatever text field the cursor is on, triggered with FN plus Control and running at roughly 450 milliseconds of latency. The second, triggered with Control plus Option, is screen-aware. Instead of transcribing what you say, it reads the current window and drafts the actual text for you, so asking it to reply to an open email produces a Gmail reply written in your voice, and asking it to follow up in a terminal produces a working prompt for Claude Code. It draws on HeyClicky's stored memory of the user to keep the tone consistent.
The launch matters for operators evaluating voice input tools because it pushes dictation past raw transcription into something closer to a screen-reading writing assistant. Early users are applying it to ad copy in Google Ads, resume editing, and everyday replies in Slack and LinkedIn. Farza has flagged that the feature is live in production and still rough in places, with active work on speed and reliability given a spike in usage after the demo.
Comments (15)
Sign in to join the discussion.
N
Nadia K.Jul 16, 2026
450ms is nice but the screen-context bit is the actual moat. Everyone can ship whisper wrappers, few can ship taste.
T
Tomas LindqvistJul 16, 2026
ok wait, dictating prompts into Claude Code is genuinely the loop I didn't know I needed. my wrists are filing a thank-you note.
P
Priya VenkateshJul 16, 2026
the demo cut was tight, but you buried the gmail reveal at the end. that should've been the first 3 seconds.
M
Marco B.Jul 16, 2026
how are you handling screen scraping on macOS? screen recording permission prompt is a conversion killer on first launch.
S
Svea H.Jul 16, 2026
what's the fallback when it misreads the screen context? does it just guess or does it ask?
K
Kwesi OtooJul 16, 2026
hot take: 'in your voice' is doing a lot of work in that tweet. show me it replying to my mom without sounding like a LinkedIn post.
J
joaquinJul 16, 2026
you can't dictate your way out of writing bad prompts, but you can dictate your way into shipping faster. this is the good trade.
R
Renée SatoJul 16, 2026
farza launches things the way most people tweet. relentless.
B
Benoit ClarkJul 16, 2026
curious about the pricing shape here. per-seat SaaS on a local-feeling mac utility is a weird sell to consumers.
I
ilyaJul 16, 2026
the voiceover pacing on the demo was suspiciously good. did you script it or dictate it with clicky? because that would be a flex.
C
Chiamaka E.Jul 16, 2026
onboarding question: how many permission dialogs before I get to say my first word? if it's more than two I'm out.
H
Harshad P.Jul 16, 2026
the tweet structure (problem, everyone does this, but we went further) is a whole genre now. works though, I read the whole thing.
L
Leila MahmoudJul 16, 2026
a dictation tool that reads my screen means it also reads my slack drafts I never sent. hope that stays local.
J
Jide OkonkwoJul 16, 2026
indie founders keep shipping features apple should've shipped in 2019. love it, keep embarrassing them.
R
Ravi TejaJul 16, 2026
the thumbnail choice was chef's kiss. clean product shot, no floating emojis, no founder face. rare restraint.
Work With Us
Book an intro call to explore our suite of services. We might have a waitlist.
heyclicky
an ai buddy that lives on your mac
an ai buddy that lives on your mac
Tags
TLVC Rating
Community Rating
No ratings yet
to rate this launch.
HeyClicky is a Mac-native AI assistant that sits next to the cursor, watches the screen, and listens for voice commands. This launch adds two dictation modes aimed at people who write across a lot of different apps, from Claude Code and terminals to Gmail, Slack, LinkedIn, and Google Slides. Founded in 2026 by Farza Majeed, who previously built Buildspace into a community of 125,000+ builders , the company is part of Y Combinator's Spring 2026 batch. The first mode is a plain speech-to-text that streams into whatever text field the cursor is on, triggered with FN plus Control and running at roughly 450 milliseconds of latency. The second, triggered with Control plus Option, is screen-aware. Instead of transcribing what you say, it reads the current window and drafts the actual text for you, so asking it to reply to an open email produces a Gmail reply written in your voice, and asking it to follow up in a terminal produces a working prompt for Claude Code. It draws on HeyClicky's stored memory of the user to keep the tone consistent. The launch matters for operators evaluating voice input tools because it pushes dictation past raw transcription into something closer to a screen-reading writing assistant. Early users are applying it to ad copy in Google Ads, resume editing, and everyday replies in Slack and LinkedIn. Farza has flagged that the feature is live in production and still rough in places, with active work on speed and reliability given a spike in usage after the demo.
Comments (15)
450ms is nice but the screen-context bit is the actual moat. Everyone can ship whisper wrappers, few can ship taste.
ok wait, dictating prompts into Claude Code is genuinely the loop I didn't know I needed. my wrists are filing a thank-you note.
the demo cut was tight, but you buried the gmail reveal at the end. that should've been the first 3 seconds.
how are you handling screen scraping on macOS? screen recording permission prompt is a conversion killer on first launch.
what's the fallback when it misreads the screen context? does it just guess or does it ask?
hot take: 'in your voice' is doing a lot of work in that tweet. show me it replying to my mom without sounding like a LinkedIn post.
you can't dictate your way out of writing bad prompts, but you can dictate your way into shipping faster. this is the good trade.
farza launches things the way most people tweet. relentless.
curious about the pricing shape here. per-seat SaaS on a local-feeling mac utility is a weird sell to consumers.
the voiceover pacing on the demo was suspiciously good. did you script it or dictate it with clicky? because that would be a flex.
onboarding question: how many permission dialogs before I get to say my first word? if it's more than two I'm out.
the tweet structure (problem, everyone does this, but we went further) is a whole genre now. works though, I read the whole thing.
a dictation tool that reads my screen means it also reads my slack drafts I never sent. hope that stays local.
indie founders keep shipping features apple should've shipped in 2019. love it, keep embarrassing them.
the thumbnail choice was chef's kiss. clean product shot, no floating emojis, no founder face. rare restraint.