A screenshot in,requirements out
Half the time the “spec” you're handed is a mockup, a legacy screen, or a screenshot in an email. Khint captures the region, reads it with a vision chat you can keep questioning, and turns the screen into draft requirements, from one keyboard shortcut, without retyping a single field.
The requirements are on the screen,
just not in words yet
A mockup or a legacy screen carries everything a requirement needs (the fields, the controls, what's required, what computes) but none of it is text you can paste into a ticket. Transcribing a screen field by field is the chore that eats an afternoon and still misses the button nobody mentioned. Khint closes that gap: Capture reads the image and lets you interrogate it, an Agent you tune once shapes the inventory into requirements, Memory keeps every screen of the feature in context, and when a draft is ready the integrations file it without leaving the page.
A screen, turned into words
One captured screen, before and after.
- Invoice # (greyed out)
- Customer: dropdown
- Line items: qty · unit price · total
- Tax % field
- Save draft / Send buttons
- The system shall auto-generate the invoice number.
- The user shall pick a customer from a dropdown.
- The user shall add line items; total is computed.
- The system shall apply a tax percentage to the subtotal.
- The user shall save a draft or send the invoice.
Illustrative: your draft follows whatever requirement style you wrote into the Action prompt.
One screen, start to a draft
A handed-over mockup, step by step.
Start a session for the feature you're speccing
Open Memory and start a session: “Invoicing rebuild”. Every capture and Action you run is logged and Khint keeps a compact running summary, so as you walk through screen after screen the model keeps the whole feature in context. How Memory works →
Capture the screen you were handed
You got a Figma export of the new “Create invoice” screen, or a screenshot of the legacy one nobody documented. Open the palette (Cmd+Shift+K), pick Capture & ask, and drag a box around it. The image lands pinned in a chat window. An emailed PNG or a phone photo works too: drop it into the Capture tab instead of dragging a box.
Ask the vision chat what it sees
Ask in plain language: “List every field, control, and button on this screen, and flag anything that looks required or auto-filled.” The screenshot stays in context, so you keep going (“What validation does the tax field imply?”, “What states does the Save button have?”) until you have the screen fully inventoried, no re-uploading between questions.
Turn the inventory into requirements
For a text-heavy screen, hit Extract text instead (the labels land on your clipboard), then select them and run a saved Agent: “rewrite these screen fields as numbered functional requirements, one per line, each starting with ‘The system shall’.” Describe the Action once and let the editor Generate the prompt; every screen after gets the same shape.
File the draft where the work lives
With Jira or Confluence connected, the palette can file the result: a Create issue row drafts a ticket from the requirements, or Create page publishes the full screen spec to Confluence, and hands the key or URL back to paste into the thread. Every capture and Action is in your Memory session too, so the spec reads back as a record of which screens you mined. How integrations work →
Reading the screen is just the first step
The drafted requirements are reusable text: paste them into a ticket, feed one line into a follow-up Action that writes acceptance criteria, or chain it. Build a workflow of up to five steps in series: an Agent that inventories, a second that writes the requirements, then a create-issue or create-page step at the end. And because Khint ships an MCP server, Claude Desktop or Codex can read the speccing session directly (the screens you captured, the requirements you drafted, the tickets you filed) without copy-paste.
Common questions
How does Khint pull requirements out of a screenshot?
Open the palette with Cmd+Shift+K and pick Capture & ask. Drag a box around the mockup or legacy screen and the image lands pinned in a chat window, where you can ask the vision model to list every field, control, and state it sees, and keep asking follow-ups in the same conversation, because the screenshot stays in context across turns. For a screen that is mostly text, Extract text instead drops the plain text on your clipboard, and you run a saved requirements Action on it. Either way the structured draft lands where your cursor is, ready to refine.
What if the screenshot was emailed to me, not on my screen?
Drop the file into the Capture tab. PNG, JPG, WebP, HEIC, GIF, BMP, and TIFF all go through the same flow as a live region capture, so a phone photo of a whiteboard, a designer's exported mockup, or a screenshot a stakeholder pasted into an email all work the same way. You get the same plain text out, ready for any Action.
Does it remember the feature across several screens?
If a Memory session is active for the feature you are speccing, the running context rides along with each Action. Capture the login screen, then the dashboard, then the settings page. Khint logs what you run and keeps a compacted summary, so by the time you draft the acceptance criteria the model already knows the screens you walked through. Sessions are stored locally in SQLite on your machine and are never synced to the cloud.
What does Khint send to the AI, and where does the image go?
Captures go to the vision model so it can read the image; Actions send only the text you selected, plus (if a Memory session is active) the compacted session summary. Nothing else leaves your device silently. Khint's free tier covers 10 AI actions and 5 captures per day with no credit card; paid plans start at €7/mo, with Pro at €29/mo for 100 AI actions a day; every plan includes all the work integrations.
Try it on the next screen you're handed
Free with 10 AI actions and 5 captures per day. No credit card. Capture the screen, ask the vision chat, keep the requirements.