The AI agent that lives beside your cursor.
Kanessa hears you, watches your screen in real time, and quietly takes the work off your hands - from drafting emails to navigating apps you've never seen before.
No credit card · Free during beta · Local-first audio & screen processing
Hey - just went through the Q4 deck you sent over. Overall it's looking sharp, but I had a few thoughts before the exec review on Thursday.
The revenue slide (slide 7) feels a bit dense. Could we split the regional breakdown into its own slide? Also the projection chart needs updated Q3 actuals.
Could you send the revised version by EOD tomorrow? Happy to jump on a quick call if easier.
Thanks,
Sarah
Drafting reply...
Works alongside the apps you already use
Windows
Chrome
Slack
Notion
Figma
Gmail
Linear
VS Code
SpotifyWindowsChromeSlackNotionFigmaGmailLinearVS CodeSpotifyCapabilities
One agent. Three senses. Infinite tasks.
Hears you, naturally.
Push to talk, wake word, or always-on. Kanessa understands context, tone and intent - no rigid commands.
Sees your screen, in real time.
Continuous vision of whatever you're looking at - apps, PDFs, dashboards. It knows what you mean by 'this'.
Acts on your laptop, end-to-end.
Clicks, types, navigates, fills forms, opens apps. Hand off the boring tasks and watch them get done.
How it works
From a sentence to a finished task.
Kanessa is built around a tight loop of listen → see → reason → act. Every step is observable, interruptible, and stays on your machine by default.
- 01Speak naturally
Hold a key, say a word, or just start talking. Kanessa listens with low-latency on-device speech.
- 02It sees what you see
Screen frames stream to the agent so it understands the context of your request - the actual pixels in front of you.
- 03It plans the steps
Kanessa reasons about the task, breaks it into clicks, keystrokes and tool calls, then asks before anything destructive.
- 04It does the work
Watch the cursor move. Take over any time. Get a summary of what changed when it's done.
Guides you through any app
Stuck in Figma or Photoshop? Kanessa walks you through it - step by step.
It watches what's on your screen, points at the right tool, and narrates the next move. Like having a senior designer over your shoulder.
Showcase
Things people say to Kanessa.
Real prompts. No setup, no scripts, no integrations to configure first.
"Reply to Sarah and say I'll send the deck by Friday."
"Find a 30-min slot with the design team next week."
"Turn the spec on my screen into a Linear ticket."
"Clean up my desktop and group these into folders."
"Show me how to set up a hero section frame in Figma."
"Hey Kanessa, teach me how to use this new software to edit videos."
FAQ
Questions, answered.
Is Kanessa always listening and watching?+
Only when you want it to be. You can run it push-to-talk, with a wake word, or always-on. Screen capture is paused by default and only activates per session - with a clear on-screen indicator.
Where is my data processed?+
Audio and screen frames are processed locally where possible. When a task requires a cloud model, only the minimum necessary context is sent, encrypted in transit, and never used to train models.
Can it actually click buttons in any app?+
Yes. Kanessa controls the OS the way you do - via accessibility APIs and pixel-aware vision. It works in browsers, native apps, and even unfamiliar interfaces.
What if I want to take over?+
Move your mouse or hit Escape and Kanessa instantly pauses. It will summarize what it has done so far and wait for instructions.
Which platforms are supported?+
Windows is in private beta today. A Linux build is on the roadmap.
Give your cursor a brain.
Join the private beta. We're onboarding new users every week.
We'll never share your email.