The workflow reads the sprint,you read the summary
Most AI shortcuts need you to paste something first. This one does not: the first step of the workflow is a Jira read, so you press the shortcut with nothing selected, Khint pulls the sprint itself, an agent turns the issue list into a written update, and the last step posts it where the squad reads.
You already know
what the board says
Monday morning, sprint 24 half spent. You open the board, scan twenty issues, work out what actually moved, and retype it as a paragraph for the people who will not open Jira. The reading is mechanical, the writing is the same shape every week, and both happen before anyone has said anything worth capturing. A Khint workflow can start from the board instead of from you: the read is a step, not a chore.
A real workflow run: the steps execute one after another and the result lands in Slack, in a channel other people are already reading.
From the board to the channel
Building the Monday sprint update as a workflow you press once.
Connect Atlassian once
On the Integrations page, open the Atlassian card and paste three things: the email on your Atlassian account, an API token, and your cloud domain. They go into your operating system keychain, not a Khint server. The palette immediately gains its Jira rows, and the workflow editor gains the Jira steps, including the three that read: get one issue, run a search, or pull a sprint.
Make the read the first step
In Workflows, start a new workflow and set step one to the Jira sprint read, pinned to the board your squad works from. Prefer a query? Use the search step and pin a JQL instead, which is how you scope the update to one epic, one label, or one assignee. Not fluent in JQL? Khint has an AI JQL generator in the palette to write the query for you.
Press the shortcut with nothing selected
This is the part that is different. A Khint workflow normally runs on the text you highlighted, but a workflow whose first step is a read does not need any: Khint knows step one supplies its own input, so an empty selection is fine. Give the workflow its own keyboard shortcut and the Monday update starts from any app you happen to be in, without you finding a document first.
Let an agent turn issues into sentences
What the read returns becomes the running text, and the next step reads it. So step two is one of your own agents: “Write the sprint update”, tuned to your house format, what closed, what slipped and why, what needs a decision this week. No prompt yet? Type a description in the agent editor and press Generate. Only this step calls the model; the read before it is an API call to Jira.
End the workflow where people read
The last step writes the update out: a Slack message to the squad channel, a Confluence page for the record, or both at once with a parallel step that dispatches up to four writes together. The run ends with the receipt, the posted message or the page URL, and the same link lands on your Home history with an Open button, so the update is one click away from the app that wrote it.
Read, write, post
Up to five steps in series. The first one supplies the input the rest work on.
Swap any step for yours. Read a JQL search instead of a sprint, put a second agent after the first to cut the update down to five lines for the exec channel, or replace the Slack write with a Confluence append so every week lands on the same running page. The shape holds: what each step returns is what the next step reads.
What it does not do
The honest edges, so the workflow behaves the way you expect on Monday.
It runs when you press the key
Khint has no scheduler. The workflow fires on its shortcut or from the palette, so the update is written when you decide to write it, not at 08:00 whether you are there or not.
A read cannot be a parallel branch
The parallel step exists to write to several tools at once, and every branch has to report an address it wrote to. Reads run as their own step in the series, before the agent that uses them.
The read costs no credits
A Jira step is a call to Jira, not to a model. Only the agent steps in the workflow are billed, so a read-heavy workflow is cheaper than its step count suggests.
A missing session never stops the run
When a Memory session is active the read is logged to it, but that logging is best effort. If no session is open the workflow still runs to the end, because its job is to feed the next step.
When the update is about one ticket, not the whole sprint
The same idea works one issue at a time. The get-issue step can take its target from the running text instead of a pinned key, so a workflow that starts from a selection reads whichever ticket key or pasted Jira link it finds there, then hands the issue to an agent. That is the escalation version of this page: select a line in a thread, press the shortcut, and get the ticket read back to you in plain sentences before you answer. It pairs with summarizing a long Jira thread, which does the same for the comments underneath.
Common questions
Can Khint read my Jira sprint without me pasting anything?
Yes. A Khint workflow is a short series of steps, and a step can be a Jira read: pull a sprint, run a JQL search, or fetch one issue. When the read is the first step, the workflow supplies its own input, so you press its keyboard shortcut with nothing selected in any app and the run starts from the board. What the read returns becomes the running text that the next step works on, which is how an AI agent placed after it ends up summarizing real issues rather than something you retyped.
How is this different from asking a chatbot to summarize my sprint?
A chatbot cannot see your board, so you export or copy the issue list in front of every prompt, then copy the answer back out to wherever it needs to go. In Khint the read, the writing, and the posting are three steps of one workflow behind one shortcut: Jira is read directly through your own API token, the agent prompt is saved once and reused every week, and the final step writes the result into Slack or Confluence so the run ends with a posted message instead of text in a chat window.
Does the sprint summary post itself automatically every morning?
No, and that is deliberate. Khint has no scheduler: workflows run when you press their shortcut or pick them in the palette. The saving is in the reading and the retyping, not in removing you from the loop. If you want a look before anything is published, end the workflow at the agent step and paste the update yourself, or keep the write step and check the receipt the run returns, which is also logged to your Home history with a link to the message.
What does Khint need to read Jira, and where do the credentials live?
Three values on the Integrations page: the email on your Atlassian account, an API token you generate in Atlassian, and your Jira cloud domain. They are stored in your operating system keychain on your own machine, and Khint talks to Jira directly from the app with them. Turning the integration off pauses it and keeps the credentials; Forget credentials clears them. Khint uses API tokens rather than an OAuth app, so nothing about your Jira account is delegated to a Khint server.
Can I build this workflow without assembling the steps by hand?
Yes. The Workflows page opens on a chat band: describe what you want, for example read the current sprint, write a status update, post it to Slack, and Khint builds the workflow on the rail underneath, reusing agents you already have and authoring the ones you do not. Integration steps arrive with their fields waiting for you, so you pick the board, the channel, or the space once in the editor and every later run replays that choice.
Let the board write its own update
Free with 300 credits a month, about 10 AI actions a day. No credit card. Build the workflow once, press one key every Monday.