Add to a Confluence page,never over it
Status pages, running specs, and team FAQs die when updating them means opening Confluence, finding the page, and pasting carefully around everyone else's edits. Khint shapes your update with a saved AI action and appends it to the end of the page you pick, from whatever app you're in, with a version-locked write that fails politely rather than overwrite a colleague's work.
The living page problem:
updating is riskier than creating
Creating a fresh Confluence page is easy; that flow is covered in meeting notes to a Confluence page. The pages that matter most, though, are the ones that already exist: the weekly status page stakeholders actually read, the spec that grows a decision at a time, the team FAQ. Those pages are shared, which is exactly what makes updating them stressful: paste in the wrong place, publish over someone else's draft edit, and a page the whole team trusts is quietly wrong. So updates wait “until I'm at my desk”, and the page drifts out of date. Khint's Confluence integration is built for that exact write: Append to page adds your block at the end and nowhere else, and Comment on page stays out of the body entirely.
One status page, kept current, step by step
A week of scattered signals, one page stakeholders can trust at the end of it.
Connect Atlassian once
On the MCP page, flip on the Atlassian card and save your email, API token, and cloud domain; they stay on your machine. The palette's Confluence section gains Append to page out of the box, and you can switch on Comment on page, Pull page, and the rest per row from the same card.
Start a session for the week's page
Open a Memory session named for the work: “Checkout revamp: weekly status”. Every Confluence pull, append, and comment is logged into it, and the session's compact running context rides along on your AI actions, so Friday's update is written knowing what Monday's said. How Memory works →
Shape the update where it happened
The status change arrives as a Slack thread, a stand-up remark, a test result. Select the raw text right there, hit Cmd+Shift+K, and run a saved Action like “Status update entry” that returns a dated, scannable block. No prompt yet? Describe it in one line in the Action editor and press Generate.
Append it to the living page
Pick Append to pageunder Confluence, choose the space, then the page (your last-used space is floated; pasting the page's URL also works). The page is chosen before the AI runs, so cancelling costs nothing. Your block lands at the end of the page as a new version; everything already on the page passes through untouched.
Chain it for the recurring update
In the Workflow editor, add your “Status update entry” Action and an Append to Confluence pageintegration step pinned to the status page. Raw notes in, formatted block appended, one keystroke. Reviewing someone else's spec instead? Run Comment on page and your feedback lands as a footer comment without touching their content. How workflows are built →
Your block lands at the end, everything else stays put
What an append looks like on a weekly status page.
The grey sections are the page as your team left it, and Khint treats them as untouchable: the existing content is carried through exactly as it was, so macros, tables, and layouts survive, and no model is ever asked to reproduce your team's writing. The blue block is the only new thing, and its shape is whatever your saved Action returns: Improve tightens the template in place as the page matures. The write itself publishes a new page version locked to the one Khint just read, which is what makes it safe on a page other people edit.
Why not let AI edit the whole page?
Because a shared page is a record, and a model that round-trips the whole record will eventually lose part of it: a macro it didn't understand, a table it flattened, a paragraph it “improved”. Khint deliberately doesn't offer that. Appends add, comments annotate, and rewriting is reserved for text you selected yourself, the flow shown in Confluence page cleanup. The same pattern composes across your stack: teams on Notion get the equivalent running-page flow in the product decision log use case, and everything Khint pulls or writes lands in your Memory session, so the update you append on Friday is written with the week's context behind it.
Common questions
What happens if a teammate edits the page while Khint is writing?
The write fails, safely. Confluence has no true append operation, so adding to a page means reading it and writing it back as a new version. Khint locks that write to the version it just read: if anyone published an edit in between, Confluence rejects Khint's version instead of letting it overwrite theirs, and Khint tells you someone edited the page while you were writing. You run the append again on the fresh version. The failure mode is a retry, never a colleague's paragraph silently disappearing.
Does the AI rewrite what's already on the page?
No, and that's deliberate. The existing page content is carried through byte-for-byte: Khint never parses it, reformats it, or asks a model to reproduce it, so macros, tables, and layouts survive exactly as they were. Only the new block you're adding is AI-shaped. If what you actually want is to rewrite a messy section, that's a different move: select the section and run a rewrite action, as shown in the Confluence page cleanup use case.
When should I comment instead of appending?
Comment on page posts a footer comment under the page rather than touching its body, so it physically cannot corrupt the page: no version, no read-modify-write, just an addition to the discussion. Use it for review feedback, open questions, or a note the page owner should triage before it becomes content. Append to page is on by default once Atlassian is connected; Comment on page is switched on per row from the Atlassian card on the MCP page, so your palette only carries the moves you use.
How does Khint know which page to write to?
You pick it before any AI runs. The palette flow opens a space picker (your last-used space floats to the top), then lists that space's twenty most recently modified pages; pasting a page URL or id works as the escape hatch. Because the page is chosen first and the pickers are free, backing out never costs you an AI action. In a workflow, the step pins a page once, so a recurring update always lands on the same page without re-picking.
Keep the page your stakeholders actually read current
Free with 10 AI actions and 5 captures per day. No credit card. The next update lands on the page before you switch windows.