Ask AI about anythingon your screen
Half of a PO or BA's day is spent in front of screens somebody else built: a legacy app mid-elicitation, a dashboard in a steering call, a diagram the engineers shared. Khint's Capture & ask pins a screenshot of it in a chat, so you can question the screen, and keep questioning it, instead of guessing what it means.
Some screens don't need copying,
they need explaining
OCR gets the words out of a screenshot, and Khint does that too, with Extract text. But the screens that slow product people down are rarely a copying problem. A 40-field legacy form during requirements elicitation, a retention chart you're asked to react to live, an unfamiliar error dialog a user just sent you: the text is right there; the meaning isn't. Capture & askis the other half of Khint's Capture pillar: instead of extracting the pixels, you interrogate them, in a chat that keeps the image in view the whole time.
From “what am I looking at?” to a written answer
A five-step walkthrough of questioning a screen you didn't build.
Frame the part of the screen you're puzzling over
Open the palette with Cmd+Shift+K and pick Capture & ask. Drag a box around the region: the legacy form, the metrics panel, the error dialog. On macOS the selection is handled by the system's own screenshot UI, the same crosshair you already know. How Capture works →
Ask it a question in plain language
The screenshot lands pinned in a chat window. Ask what you'd ask a colleague who built the thing: “which of these fields look mandatory?”, “what is this chart saying about activation?”, “what does this error usually mean?” The reply comes back as plain text: no formatting to clean up before it goes in your notes.
Drill down without re-uploading
The image stays in context across turns, so the conversation can move from understanding to output: “explain this screen”, then “now list the business rules it implies”, then “phrase those as three acceptance criteria”. One capture, as many follow-ups as the screen deserves. Extract requirements from a screenshot →
It knows what you're working on, and remembers what it told you
With a Memorysession active, the chat sees the compacted context of your current work, so its answers land in your project's vocabulary. And every question and answer is logged back into that session and your local history: the explanation you got at 10am is still there when you write the ticket at 4pm.
Turn the answer into work
The reply is ordinary text, so it feeds anything downstream: paste it into your spec, run a saved Action to reshape it, or, with your tracker connected on the integrations page, push it out to Jira, Confluence, or Linear from the same palette. Capture an error message into a ticket →
Why not just paste the screenshot into a chatbot?
You can. It's just a longer trip. Screenshot to disk, switch to the browser, find the right thread, upload, ask, copy the answer back, and now your question about an unreleased screen lives in a chatbot account's history. Capture & ask collapses that into one gesture from the palette you already use for Actions, and it's wired into your workday on both ends: the chat reads your active Memorysession's context before answering, and writes each exchange back into that session, which stays in a local database on your machine rather than in someone else's cloud thread. Asking about a screen becomes part of the same flow as acting on it.
Common questions
How is Capture & ask different from Extract text?
Extract text is the one-shot grab: drag a box, the OCR'd characters land on your clipboard, done. Capture & ask is the read-and-reason path: the screenshot pins in a chat window and stays in context, so you can question it across several turns: what a chart implies, which fields on a form are required, what an error actually means. Reach for Extract text when you need the words; reach for Capture & ask when you need the meaning.
Do I need an API key or any setup to use it?
No. Capture & ask works out of the box through Khint's vision service: you don't bring a key or configure a model. The free plan includes 5 captures per day, and that allowance is separate from your AI Action quota, so questioning a screen doesn't eat into the actions you save for writing tickets.
Does Khint watch my screen in the background?
No. Nothing is captured until you pick Capture & ask from the palette and drag a box yourself. On macOS the region selection is done by the system's own screenshot UI: Khint's process never calls a screen-capture API and doesn't ask for the Screen Recording permission. One trigger, one region, one screenshot: what you framed is what the chat sees.
Do the answers land anywhere I can reuse them?
Yes. Every exchange is written to your local history, and when a Memory session is active each question and answer is also logged as an entry in that session, so what you learned from the screen rides along into your next Action. Sessions live in a local SQLite database on your machine and are never synced to Khint's servers. The chat is session-aware in the other direction too: with a session active, it sees the compacted context of what you've been working on, so it answers a screenshot question knowing which feature you're in the middle of.
Stop guessing what a screen means
Free with 10 AI actions and 5 captures per day. No credit card. Capture & ask lives in the palette, one shortcut away.