Three reads, three writes
Reads land in your active session; writes go straight to your site. The defaults are curated rather than everything-on: a palette showing six Jira rows on day one is a palette nobody reads.
IntegrationsJira
File a Jira issuefrom wherever you are
Select text in any app, hit the palette, and Khint writes the issue and files it. The other direction works too: pull an issue, your active sprint or a JQL search into the work session your agents already read.
KhintPalette
Sprint 24
/
Jira
Six rows, in the launcher you already open
Connecting Jira doesn't add a Jira window to Khint. It adds a section to the palette, and you choose which rows appear: Create issue and Change status are on out of the box, the four others are one checkbox away.
What connecting Jira
actually changes
Four things, and none of them is a new window to visit. The rows land in the launcher, what you pull or file lands in your work session, the same writes are available to a workflow, and switching the card off keeps your credentials.
Pull issuereadThe receipt goes back into the session
A created issue is logged as a session entry after Jira confirms, so the next agent runs knowing the ticket exists, and so does the AI client reading your session over MCP.
Or fan it out from one text
Create issue and Add comment are also workflow steps. A parallel step takes two to four destinations and dispatches them at once, so one press files the ticket, publishes the page and posts the update.
Post update
Publish pageOff means paused, not wiped
Turning the card off hides the rows and stops workflows using it, while the token stays in your keychain for an instant reconnect. Forget credentials is a separate, deliberate two-click button.
JiraPausedIt acts as you,
and only where you can
An integration that writes into your team's tracker has to be explicit about its limits. A personal token carries your own permissions, so there's no access model to configure and none to get wrong.
Security and privacyThe token lives in your OS keychain
Calls go from your Mac straight to your site
Khint's backend never sees it, never proxies it
It reaches exactly what your account can open
The project or issue is picked before the model runs
A landed write is never reported as a failure
Pause keeps the credentials, Forget clears them
Jira Cloud, token only: no Server or Data Center
Connect Jira in about two minutes
A personal API token, your email and your site. Nothing to install in Jira, no admin approval, no OAuth app to register.
Create an API token
Go to id.atlassian.com, open Security, then API tokens, and create one. Copy it: Atlassian shows it once.
Open Integrations in Khint and turn Atlassian on
Enter your Atlassian email, the token, and your site (mycorp.atlassian.net), then Save. The token goes into your operating system's keychain, not into a config file.
Choose the rows you want
Create issue and Change status are on by default. Pull issue, Active sprint, Search and Add comment are one checkbox away on the same card.
Use it, and watch it land in your session
Cmd+Shift+K anywhere. Everything you pull or create is logged into the active Memory session, so the agent you run next already knows about it.
- The same credential connects Confluence: one card, both tools. And once Jira is connected, Knowledge can keep a local full-text index of your issues so agents search them without a pull.
The rest of Khint
An integration is where work comes in and goes out. These three are what happens in between.
Common questions about Jira and Khint
Do I need a Jira admin to install anything?
No. You create a personal API token on your own Atlassian account and paste it into Khint. Nothing is installed into the Jira site, no app is registered, and no admin has to approve anything. The token carries your permissions, so Khint reaches exactly what you can already open.
Does my Jira data go through Khint's servers?
No. The token is stored in your operating system's keychain and every REST call goes from your Mac straight to your Atlassian site. What you pull lands in a local Memory session, which is never synced. If an agent then rewrites that text, only the text you gave it reaches the model, exactly as with any other action.
What does a Jira write cost?
The Jira call itself is free. What costs is the writing: when the row hands your selected text to an agent to turn it into an issue title and description, that's one normal agent run against your monthly credits. Pulls make no AI call at all, so they're free, and so is picking a project or an issue.
Can I use Jira inside a workflow rather than as a single row?
Yes. Create issue and Add comment are also steps a workflow can take, so one press can clean up your notes, file the issue and post the link to Slack. A parallel step can even write to Jira, Slack and Confluence at the same time from the same running text.
Can Claude Desktop see my Jira through Khint?
Indirectly, and only what you pulled. Anything a Jira row pulled into your active session is part of the context Claude Desktop or Codex read over MCP. Khint doesn't expose a Jira tool to those clients: they read your session, not your tracker.
Which Jira does this work with?
Jira Cloud, with a personal API token. Server and Data Center are out of scope, and OAuth was deliberately deferred rather than half-built. Your site is the part before /jira in your Jira URL, for example mycorp.atlassian.net.
Stop retyping your own notes into Jira
Free with 300 credits a month. No credit card. macOS 13+ or Windows 10 22H2+.