One AI toolkit forcompliance officers
Regulations into obligations. Policy drafts reviewed in place, formatting intact. Audit answers written from what your team already documented. One menu-bar app, one shortcut, and the sensitive text never sits in a chat history.
Every task starts with text
you are not allowed to paste
A compliance day is reading: an article of a regulation, a policy draft, a vendor's security answers, a contract clause. Then you rewrite it for someone else. The obvious tool is a chat tab, and a draft policy or a client's data agreement is exactly what you cannot drop into one. Khint runs the same pass where the text already lives: an agent on what you selected, and nothing else from your screen.
The palette comes to the text, not the other way around
No window to open, no tab to switch, nothing to paste twice. Khint reads the selection where it is and writes the answer back in the same place.
The real palette: an agent runs on the selected text and the rewritten version lands back in place.
A gap-assessment day,
clause to control
Every step below is a shipped feature, not a roadmap slide.
- 09:10Open a session for the assessmentPress Start on a Memory session called 'DORA gap assessment'. From there, every action you run is logged to it and compacted into a short running context that later actions can see, so you stop restating the scope at each prompt. Sessions live in a local database on your machine and are never synced to a server.
- 09:30Turn the article into obligationsSelect the article in your browser or your PDF viewer, hit Cmd+Shift+K, and run a 'Clause to obligations' agent: one line per obligation, who owns it, what would evidence it. The agent reads the passage you highlighted, not the rest of the page.
- 10:15Point an agent at the whole textFor the long ones, pin the file itself: an agent can take a Document as its input and read PDF, Word, PowerPoint, Excel and OpenDocument files. Anything past roughly 10 KB is condensed into a cached summary that is only regenerated when the file actually changes, so a 200-page directive is not re-sent on every run.
- 11:00Review the policy and get the file backAn agent whose input is a document can hand the document back instead of text. Khint sends only the paragraphs, writes back only the ones that changed, and leaves every style, table and header untouched. Your source file is never opened for writing: the edited version is saved as a new copy.
- 14:00Answer the questionnaire from what existsSwitch on Search Knowledge for the agent. Khint keeps a local index of your Jira, Confluence, Notion and Linear content, built on this machine and never on our servers, and slips the best-matching excerpts into the run. The answer quotes your own documented controls instead of inventing them.
- 16:30File the control and the remediationOne workflow closes the loop: an agent formats the control write-up, a Confluence step creates the page, and a parallel step opens the remediation issue in Jira and posts the note to the channel at the same time. The workflow ends by handing you the created page URL.
Start from packs,
not a blank prompt
The wording of an obligation is house style, so it belongs in a saved agent rather than in someone's head. Clone a starter pack such as Doc Polish to get a working review agent, then describe what you want in one sentence and press Generate: Khint drafts the prompt, and Improve refines it in place. Describe a whole sequence in plain English and the workflow generator assembles the steps for you. When a pack works, share it, and every analyst on the team clones the same version instead of writing their own. Each agent also carries a Redact PII switch: turn it on for the ones that touch client text and emails, phone numbers and account identifiers are replaced with placeholders before the text is sent.
The four pieces of the toolkit
Product overviewCommon questions from compliance and risk teams
What does a compliance officer actually use Khint for?
The reading and writing around the controls. Select an article of a regulation, a policy paragraph or a vendor's answer, press Cmd+Shift+K, and run a saved agent on it in place: turn a clause into a list of obligations, rewrite a policy section in plain language, draft the answer to an audit question, or summarise a supplier's security pack. The same palette captures any region of your screen to text, reads a pinned PDF or Word file, creates the Confluence page and opens the Jira issue, and chains up to five of those steps into one workflow.
Is Khint a GRC or compliance management platform?
No, and it does not try to be. Khint holds no control register, no evidence store and no audit trail of your programme, and nothing it drafts is legal advice: a human still reviews and signs off. Your register stays where it already is, in Jira, Confluence, Notion, Linear or your GRC platform. Khint is the layer that writes into them, so the work of turning dense text into something filed takes a keystroke instead of an afternoon.
Where does a draft policy or a confidential clause go?
An agent only ever sees the text you selected or the file you pinned, plus the compact summary of your active Memory session if you opted that agent in. That goes to Khint's proxy for the model call and is not kept. What stays on your device stays there: Memory sessions and their entries live in a local SQLite database that is never synced, the input and output text of your history never leaves the machine, the Knowledge index of your Jira and Confluence is built locally, and integration credentials sit in your OS keychain. For the agents that touch client data, the per-agent Redact PII option swaps emails, phone numbers and account identifiers for placeholders before anything is forwarded.
Can it read a 200-page regulation or a scanned PDF?
A long text-based PDF, yes: pin it as an agent's document input and Khint condenses it into a cached summary, regenerated only when the file changes. A scanned PDF is the honest limit. If the file carries no text layer, Khint tells you so rather than returning an empty answer, and you can capture the pages on screen instead: Extract text runs OCR on any region you drag, and Capture and ask lets you question the screenshot directly.
Is there a free plan, and does it run on Windows?
Yes to both. The free tier includes 300 credits a month, about 10 AI actions a day, with no credit card. Paid plans start at €7 a month, Pro is €29 a month for 4,000 credits, and every plan includes the same work integrations. Khint runs on macOS 13+ and Windows 10 22H2+.
Run your next gap assessment with it
Free with 300 credits a month, about 10 AI actions a day. No credit card. The sensitive reading stays on your machine, and the filing takes one keystroke.