Kipwise Agent does the work your team keeps asking each other to do in Slack: find the document, read the page, write it up, check the code, pull the numbers, answer the customer. Mention it once and every reply in that thread is a conversation with it.
Knowledge gets written down once and asked about forever. These are the threads that used to cost someone an hour.
Ask in the thread. It searches your knowledge base, reads the page and answers with the relevant part and a link to the source. It only ever sees pages you can see, so an answer is never drawn from something you were not allowed to read.
Tell it to write it up. It drafts the page from what was said, files it in the folder you name, publishes it and posts the link. It appends to an existing page rather than creating a duplicate, and it says so when one already covers the topic. Your name goes on it, because it wrote it as you.
It lists the comments on a page and posts your reply, surfaces the open questions in your team’s Q&A so you can answer one by linking the page that settles it, shows the pages waiting on your review, and reminds you which must-reads you still have to confirm.
It reads the actual file, commit or pull request and answers with the lines it read, linked. Ask it to change something and it opens a pull request on its own branch. It never merges. Code questions run on the strongest model available, so the answer is worth the wait.
Ask for top queries, clicks, positions or active users over a period. It pulls the numbers from Search Console and Analytics and states the date range next to every figure. It reports what it actually fetched, never a guess.
It reads the conversation, finds and reads the help center article, and drafts an internal note. Sending anything a customer can see is a separate, explicit step that only people you have approved can trigger.
Ask it to run a request later, or every week. It posts the result in the channel at that time, with exactly the same tools and permissions as a live request from you.
A CSV, a log file, a quick calculation. It runs the code in a throwaway, network-isolated sandbox tied to your thread, reports the result, and the sandbox is discarded afterwards.
The agent has no privileged account. When it reads or writes in Kipwise it does so as your Kipwise user. When it opens a pull request it uses the GitHub access you connected. When it reads a file in Drive it uses your Google access.
AI assistants are easy to adopt and hard to account for. Every model call the agent makes is logged with its tokens and its cost in dollars, attributed to the person whose request it served.
An agent you can point at internal documents is only useful if you can predict it.
One setup fee. Half of it comes straight back to you as credits.
$1,000one-time setup
Mention the agent in a channel it has been added to and ask in plain language. Reply in the same thread to keep going — you do not need to mention it again.
Anyone in a Slack workspace where it is installed. What each person can reach through it is exactly what they can reach directly.
Kipwise works immediately — if you are in the workspace, you are set up. GitHub and Google need a one-time connection, and the agent offers the link when a request needs it. Only you can complete it.
Kipwise search, pages, folders, comments, Q&A, reviews and must-reads. GitHub code, commits and pull requests. Google Drive, Gmail and Calendar. Search Console and Analytics. Intercom. A code sandbox. And scheduling. Your workspace can enable or restrict any of them.
Deleting, moving, sharing and approving each require an explicit request for that specific item, and page deletions can be restored. Edits keep the page’s existing structure, and the previous version stays in the page history.
Ordinary requests start on a fast model and step up to a stronger one when the request turns out to be hard. Code and engineering requests run on the strongest model from the start.
Yes. It talks to the region your workspace lives in, and data does not cross regions.
There is a per-person rate limit so one runaway thread cannot consume the whole team’s budget. Normal use never reaches it.
Tell it in the thread. It re-reads and corrects itself, and it prefers saying “I could not find that” to inventing an answer.
Yes — that is what the setup fee covers. Teams define their own personas, each with its own focus and instructions, and the agent picks the right one for each request.
