One AI toolkit forrelease managers

The readiness gate, the release notes, the go-live announcement, the postmortem. Four documents everyone needs on the same afternoon, written from the real board state instead of from memory, behind one shortcut.

Free 300 credits a month, about 10 AI actions a day · Plans from €7/mo · macOS 13+ · Windows 10 22H2+

You coordinate the release,
then you retype it four times

A release window is not one document, it is the same facts rewritten for four audiences in a day: a gate for the decision-makers, notes for the users, an announcement for the channel, a postmortem for the team. And the facts live in a board, a pipeline and a chat thread that none of those documents can read.

Khint puts the whole window behind one shortcut. Pull the real scope out of Jira into a work session, so every draft after that is written from the board rather than from memory. Keep the readiness gate and the release notes as saved Actions, so the format is identical whoever cuts the release. Then announce it to Confluence, Slack and Jira in one keystroke instead of three copy-pastes, each with its own chance to ship the stale version.

A release window,
freeze to postmortem

Every step below is a shipped feature, not a roadmap slide.

  1. D-2Open one Memory session for the whole windowPress Start on a session called 'Release 2.4 go-live'. From here to the wrap-up, every pull, draft and publish lands in it, compacted into a short working state that later Actions read. The release stops living in your head and in six browser tabs.
  2. D-2Pull the real scope, do not reconstruct itFrom the palette, pull the active sprint or run a JQL search for what is actually going out. Cannot remember the JQL syntax? Describe the query in plain English and Khint generates it. The result lands as an entry in the session, so the next four drafts are working from the board, not from memory.
  3. D-1Draft the go/no-go readiness gateSelect the pulled scope and run your readiness Action: open defects, untested stories, missing sign-offs, laid out in the same gate format every release. Publish it straight to Confluence from the palette so the go/no-go call has a page to sit on instead of a chat thread.
  4. D-1Turn closed tickets into user-facing release notesTicket titles are written for developers. Select the closed list and run your release-notes Action to get the user-facing version: grouped by theme, jargon translated, in your house tone. Same Action every release, so the notes read consistently even when a different person cuts them.
  5. D-0 09:00Announce once, land in three placesRun the go-live workflow: an agent shapes the announcement, then one fan-out step writes it to the Confluence release page, your Slack channel and the Jira epic at the same time. One keystroke instead of three copy-pastes, each with its own chance to post the stale version.
  6. D-0 11:20A deploy check fails: capture the evidenceA pipeline log or a monitoring dashboard shows an error you cannot select or copy. Extract text pulls the exact message and codes off the screen to your clipboard, and Capture & ask lets you question the screenshot when you need a read on what broke before you escalate.
  7. D-0 11:35Raise the blocker and move the boardDraft the incident ticket from what you captured and create the Jira issue from the palette. When the fix ships, Change status walks you through your recent issues and the transitions your workflow actually allows, so the board is honest without a trip to the browser.
  8. D+1Write the postmortem while it is still trueThe session already holds the timeline: what was pulled, what was cut, when the blocker was raised, what was announced. Run your postmortem Action against it, publish the page, then Copy recap for the release retro. Nobody spends the morning reconstructing yesterday from Slack scrollback.

The announcement is one text,
so send it once

A go-live note that exists in three places by copy-paste is three documents that will drift by the first hotfix. In a Khint Workflow, one step can fan out to between two and four integration writes that all receive the same text: append to the Confluence release page, post to the Slack channel, comment on the Jira epic. They are dispatched together, and the result of the step is one line per destination, so an agent placed after it reads the receipts (the page that was updated, the issue that was commented) rather than the draft. If one destination fails, the ones that already published stay published: a Slack message cannot be un-posted, so Khint tells you which branch missed instead of pretending the whole thing failed and inviting you to send it all twice.

Build the release pack once,
every release cuts the same way

Keep your readiness gate, release notes, announcement and postmortem Actions together in one pack, and the palette shows exactly that set while the window is open. You do not have to write the prompts from scratch: type a one-line description in the Action editor and press Generate, then tighten it with Improve. Describe the routine in plain English, for example "draft the go-live note from these tickets, then post it to Slack and Confluence", and the AI workflow generator assembles the chain with the integration steps wired in. When the pack works, share it: the person covering for you next sprint joins with a link, previews every prompt before cloning, and cuts the release exactly the way you do.

Common questions from release managers

What does a release manager actually use Khint for?

The release window is mostly writing the same four things under time pressure: the readiness gate, the release notes, the go-live announcement, and the postmortem. Khint keeps each of them as a saved Action behind one shortcut, so you select whatever raw material you have (a sprint dump, a list of closed tickets, your notes from the go/no-go call) and get the drafted artifact in the format your organisation expects. With Jira, Confluence and Slack connected, the palette also files and publishes, so you are not tabbing between three tools while the deploy is running.

Can one shortcut announce a release in several places at once?

Yes. A workflow step can fan out to between two and four integration writes that all receive the same text: append the note to the Confluence release page, post it to your Slack channel, and comment on the Jira epic in the same run. The writes are dispatched together, and anything that lands stays landed: if one destination fails, Khint reports that branch rather than rolling back the ones that already published, because a message posted to Slack cannot be un-posted.

How does a Memory session help across a multi-day release?

Start one session for the window, for example 'Release 2.4 go-live'. Every pull, draft and integration write you make lands in it, and Khint keeps a compacted working state (objective, open threads, decisions, artifacts) that later Actions can see. So the postmortem you draft on Thursday already knows about the scope cut you made on Monday. At the end, Copy recap turns the session into paste-ready markdown for your wrap-up, and Claude Desktop or Codex can read the same live session over the bundled MCP server.

Can Khint read the state of the release out of Jira?

It can pull, not watch. From the palette you can pull a specific issue, pull the active sprint, or run a JQL search (and if you do not write JQL, describe the query in plain English and Khint generates it). The result lands in your active Memory session as an entry, which is what makes the next draft factual instead of remembered. It is a pull you trigger, not a background sync: nothing is polled and no board is monitored.

Do release notes and incident text leave my machine?

Only what you act on, when you act on it. An Action sees the text you selected plus, if you opted in, your active session's compact summary. Memory sessions live in a local database on your machine and are never synced to a server, your Jira, Confluence and Slack credentials stay in the OS keychain, and an Action can opt in to PII redaction so emails and phone numbers are replaced with placeholders before the text reaches the AI. Every integration write is confirmed by you before it happens.

Cut your next release with it

Free with 300 credits a month, about 10 AI actions a day. No credit card. Connect Jira, Confluence and Slack when you are ready: it takes an API token, not an admin.