Describe the Jira search,Khint writes the JQL
JQL can answer almost any question about your backlog, if you remember the field names, the operators, and where the quotes go. Khint's palette takes the question in plain English, generates the JQL, shows it to you before anything runs, and pulls the results into your work session.
The answer is in Jira,
the syntax is in the way
Every PO knows the moment: someone asks “how many bugs are still open from the last release?” and you know Jira knows, but the basic search won't express it and the JQL bar wants statusCategory, -2w, and startOfWeek() spelled exactly right. So you click through filters, give up halfway, or ping the one colleague who speaks fluent JQL. Khint puts the question back in your language: describe the search, approve the generated query, and the results arrive where your next piece of writing happens, inside your Memory session.
Plain English in, tickets out
Finding two-week-old payment bugs in the middle of a refinement session.
Connect Jira once, switch on Search
On the MCP page, open the Atlassian card and enter your account email, an API token, and your cloud site domain: they go straight into your OS keychain. Palette rows are curated so writes are on and reads are opt-in, so flip the Search row on from the same card. It now appears in your palette under Jira.
Describe the search in words
Mid-refinement, you need the payment bugs nobody has touched. Hit Cmd+Shift+K, pick Jira Search, and choose describe it in words: “open bugs in the payments project that haven't been updated in two weeks”. No operators, no field names, no quotes to balance.
Approve the JQL before it runs
Khint answers with the generated JQL and waits. You read the query, and only your confirmation runs it: a wrong guess costs you a cancel and a rephrase, never a mystery result set you silently trusted. As a side effect, you see correct JQL for your own real questions every day, which is how the syntax finally sticks.
The results land in your session
The matching issues are pulled into your active Memorysession as a Jira-search entry, on your machine. Khint folds them into the session's compact summary, so the Actionyou run next (“draft the refinement note”, “write my status update”) already knows which tickets came back without you pasting anything.
Repeat searches get shortcuts
The same Search menu keeps two one-click presets for the everyday pulls: my open issues and updated this week. When you already know the exact query, a raw JQL mode takes it verbatim. And the separate Active sprintrow pulls a board's current sprint without any query at all.
What the translation looks like
Illustrative examples: the question you type, and the kind of JQL you approve.
The preview step is the point. A query language you never see stays magic, and magic is hard to trust with your backlog. Because Khint shows the JQL for every search you describe, you keep control of what actually runs, and you passively absorb the syntax on questions you genuinely cared about. Plenty of users will graduate to typing raw JQL for their favorite searches; the menu keeps a mode for exactly that.
Why not ask a chatbot to write the JQL?
A chatbot will happily draft JQL, and then you copy it, switch to Jira, paste it, fix the quoting, run it, and copy the results back out to wherever you were writing. Khint collapses that loop into one palette command: the query is generated, previewed, and executed in place, and the results land in the same session your next AI action reads from. That last part is the compounding win: a search feeds a status report, a grooming pass, or a stakeholder answer without a single copy-paste. And when the next step is writing to Jira rather than reading from it, the palette's create and update rows pick up where the search left off.
Common questions
Do I need to learn JQL to use this?
No. In the palette, the Jira Search command offers a 'describe it in words' mode: you type what you want in plain English and Khint generates the JQL for you. But you also see the generated query before it runs, every time, so the syntax stops being a black box: after a week of previews you will recognize the patterns, and when you eventually want to hand-write a query, the same menu has a raw JQL mode waiting.
What if the generated JQL isn't what I meant?
Nothing runs behind your back. Khint always shows you the generated JQL in a confirmation step first: if the query doesn't match what you had in mind, cancel it, rephrase your description, and run it again. For the searches you repeat every day, the same menu also has two one-click presets (your open issues, and everything updated this week) that involve no generation at all.
Where do the search results go?
Into your active Memory session, as a typed Jira-search entry. Memory sessions live in a local SQLite database on your machine and are never synced to Khint's servers. Because the session's compacted summary rides along on your next AI actions, the tickets you just pulled become context: run a 'draft my status update' action right after the search and it already knows what came back.
What does Khint need to connect to Jira?
Your Atlassian account email, an API token, and your cloud site domain, entered once on the MCP page's Atlassian card. The credentials are stored in your OS keychain, not in a config file and not on Khint's servers. Palette rows are individually toggleable from the same card, and read rows like Search are off by default until you switch them on. If you want a break, the card's toggle pauses the integration while keeping the credentials saved; a separate, explicit button forgets them entirely.
Can it change my tickets too, or only search?
Search is read-only: it pulls results into your session and touches nothing in Jira. Writing is a separate set of palette rows (Create issue, Add comment, Change status), each with its own explicit flow and its own confirmation. If the writing side is what you're after, see writing to Jira from anywhere.
Ask your backlog a question in English
Free with 10 AI actions and 5 captures per day. No credit card. Connect Jira once; search it from any app after.