Talk to kanman

When you open your workspace, you land in a conversation with kanman. kanman is the teammate that writes your stories, builds them, checks them and tells you how it went. The conversation is the one place where you see what happened, answer what needs you and hand over new work.

What you see when you open it

• What happened since your last visit. New messages sit below a line that says how many are new, for example “4 new”. The conversation opens right there.
• Which chat you are in. The top bar names it, right next to the workspace switcher: Personal or the team, with its mark.
• What needs you. kanman’s questions for you are pinned at the bottom of the chat, over the message box, one at a time; see Answer kanman’s questions. In Personal they come from every team, each labelled with its team.
• Cards. kanman writes one sentence and puts the details into a card. Every card first says what it is for, in a small line at its top: Decision, Run, Weekly report, Incident, Drafted stories, Done on my own and so on, in Personal also which team it belongs to, with its status on the right. Below follows what it is about, for example the story or the question. A card that needs your answer has an amber edge and says so in that top line. Cards stay up to date while you read them, for example an incident card shows when the incident moves from “Open” to “Mitigating”.

Personal and your team chats

You talk with kanman in separate chats: Personal is yours alone, and every team has one chat for the whole team.

• Personal is the one conversation across all your teams. It shows what concerns the teams you are a member of, plus what concerns the whole workspace (runners, billing, new members). It opens when you open the workspace, centred on the screen, with nothing beside it. Ask what needs you, how things are going everywhere, or anything that is not about one team. The short summary of what happened since your last visit (“Since yesterday: …”, or “Since Friday: …” after a weekend) arrives here, once a day when something happened, and the conversation opens at it, with a row for each team that has news: what happened there and Open chat, which takes you to that team’s chat. The team with a decision waiting for you comes first, and its button says what you do there: Answer 2 decisions. Ask “How are the teams doing?” for one sentence per team, the teams that need you first, or “Show me the weekly report” for the newest report of each team.
• Every team has its own chat, shared by the team. Everyone who works with the team reads and writes it; see A team’s chat is shared. Everything about that team lands there: its runs, decisions, reports, incidents and setup. When you open a team’s chat, the conversation moves to the left and the team’s board opens beside it. On a phone the team’s chat fills the screen; the mind button (a brain icon) next to the chip in the message box opens the team’s board and settings from the bottom. The team’s status and what needs you are in the chip’s menu.

To switch, use the chip under the message box. It lists Personal and your teams, each with how many new messages it has and how many need you. A dot on the chip tells you that something happened in another chat. The chip is the only place to switch: the conversation decides what you see. You can also press G then T to open it from the keyboard, use the arrow keys and press Enter.

Personal is private: other people in the workspace see the shared results (the board, runs, decisions and the audit log), not what you asked kanman there.

A team’s chat is shared

A team’s chat is one chat for the whole team. Everyone who works with the team reads it and writes in it, under their own name; Personal stays private, only you and kanman.

• Who wrote what. Your own messages stay on the right. Messages of other people appear on the left, where kanman writes, with their picture and name on the first message of their turn. In a team’s chat kanman’s messages carry its name too.
• Live for everyone. A message appears for everyone who has the chat open the moment it is sent.
• kanman answers when you talk to it. People also talk to each other in a team’s chat, so kanman answers only messages meant for it: a question or request to kanman, a reply to one of its messages, or a message that starts with @kanman. Messages between people get no answer and no “Noted”. When you are the only one writing, every message is for kanman.
• kanman answers once. When someone asks kanman in the team’s chat, there is one answer for everyone, addressed to the person who asked. When several people talk, kanman calls them by name.
• One language. kanman writes in a team’s chat in one language, so everyone reads the same. When the team starts, kanman takes the language its members use the app in; if they use different ones, it suggests English in the chat. Change it any time by telling kanman (“please write in German here”) or under Team > Team > Chat language. If you write to kanman in another language, it answers you in yours, and under any of kanman’s messages in another language you can choose Show in your language. Your own app language, your notifications and everything addressed only to you always stay in the language you chose.
• One message per event. kanman posts what happens on the team (runs, decisions, reports, incidents, reviews) once in the chat, not once per person.
• For one person. A message that is meant for some people says so, for example For Anna or For you and Ben. Everyone can read it, but only those people get it pinned over the message box and counted as needing them. A question everyone on the team can answer is pinned for all of them.
• Who gets notified: the team’s members, each at the level they chose for the team (see notifications). That decides who hears about a message, not who can read it.
• Who directs kanman: anyone in the workspace can read and write in a team’s chat, but kanman takes instructions only from the team’s members. If you are not a member, the message box says Join to work with kanman here; your messages stay in the chat for the team to read. See Team members.
• What stays personal. Being mentioned in a note or assigned a story is told to you in Personal, and so is anything about your account or the whole workspace.
• Before you send. When you write in a team’s chat, a line in the message box says Everyone in Sandbox Team sees this; in Personal it says Only you and kanman.

Team chats became shared on 21 October 2026. What you wrote in a team’s chat before that date stays visible only to you: you still see it in the team’s chat, above the shared messages, and nobody else does.

What kanman is working on

Between the chat and the message box, one line says what kanman is doing right now, for example Working on 2 things. Its dot breathes while kanman works and turns amber when a task waits for your answer. In a team’s chat it counts that team’s work; in Personal the work of every team. When nothing runs and nothing waits, the line folds away.

Under the line, right above the message box, kanman lists what it has running:

• every story kanman’s coding tool works on, with the story (a link to its live work log), what it does right now (“Claude Code is running go test in okta-sync”), changing in place as the work moves on, and how long it has been active;
• kanman’s own look at the team’s flow while it runs, with its newest thought;
• below them, the tasks that wait, with Answer on one that waits for you, which opens that question;
• Stop on each task, which asks once more (“Stop INT-1079?”) before it stops the run.

With up to three tasks the list is open; with more, and on a phone, only the line shows. Select Working on 2 things to open or close the list; All tasks beside it opens everything kanman works on for the team: at work now, finished and earlier.

Stories that wait on people

The questions for you are pinned over the message box, see Answer kanman’s questions, and listed in For you. Who else a story waits on shows on the story’s page.

A task whose question kanman is still settling itself says so (“I’m settling the question myself”); it does not count as waiting for you.

Watch kanman think

When kanman looks at a team’s flow on its own, its steps appear in the chat as it takes them, one line after the other: what it looks at, what it finds, what it decides. The current line is highlighted while the next one is being written. When kanman is done, the steps fold into one line that says what it did, how many steps it took and how long, for example INT-246 started · 3 steps · 41 s, or what it concluded when there was nothing to change. Select the line to read the steps again, with what the action did. Overturn stays right under it, so you can undo what kanman did on its own.

Open a card

Click a card (a run, a story, a decision, a report or an incident) to open its page. On a large screen the page opens beside the conversation, on a phone as a sheet from the bottom. Close, Escape or swiping the sheet down brings you back to the same place in the conversation. In a team’s chat, closing a page brings you back to the team’s board, never out of the team’s chat. See Tour of the app.

Answer kanman’s questions

When kanman needs you, for example to choose between two options or to grant access, the question is pinned at the bottom of the chat, over the message box, and the chat behind it dims. You see one question at a time: a small label says which story it is about (and in Personal which team), then the question, Why with kanman’s full reasoning, and the options one below the other. The one kanman recommends is marked Recommended; Add a note lets you tell kanman more with your answer.

• 1 of 3 and the arrows step through your questions. In Personal you see the questions of all your teams, newest first; in a team’s chat that team’s questions. The same question for several stories is asked once, and one answer settles it for all of them.
• Click an option: it counts at once. The question shows your answer for a moment, then the next one comes, or the panel slides away when nothing is left. Your answer stays in the chat as one line at the time you answered, for example You answered: Use the sandbox account · INT-123.
• None of the options fits? Type into Something else? Tell me… under the options and send it: kanman gets it as a reply to that question and answers in the chat.
• Answer later (or Escape) puts the questions aside so you can write a message. A small 2 questions for you above the message box brings them back, as does Answer on a story or a task. The chip under the message box shows for each team how many questions wait for you.

Questions kanman is still settling itself never appear here. You can also answer the same question in Slack or Microsoft Teams where your team uses them; whichever you use first counts everywhere.

One line per story

Everything about one story is one line in the team’s chat, in kanman’s own words, from the moment kanman writes the story to the merge:

● I’m working on INT-262. Add repository instructions to okta-sync Claude Code is editing internal/okta/client.go

The first line says where the story is: I wrote INT-262. It waits in the Backlog., I’m working on INT-262., INT-262 waits for your answer., I’m settling INT-262’s question myself., I delivered INT-262. When this changes, the new sentence fades in where the old one was. The dot in front breathes while kanman works, turns amber when the story waits for you, and becomes a check once it is delivered. The story key opens the run.

Below it, the next line says what kanman does right now and changes as the work goes on; select it to open the live work log. When the story waits for your answer, Answer opens the question. If the run stopped, the line says why. When the change goes out to further repositories, the line says how many merge requests are open and merged, with links to the ones waiting for you. While the story waits in the Backlog, Start now starts it.

History opens the rest: how far the run got (Planning, Implementing, Verifying, Review, Done), how long kanman worked on it, the run’s ID with a copy button, the steps with their time (Started · Plan approved by you · Attempt 2) and, once delivered, kanman’s debrief.

The line changes in place, so a busy team does not fill your chat with a message for every step. When something important happens, it moves to the bottom of the chat.

kanman writes a new message only when something concerns you:

• a decision that needs your answer, for example the plan or the merge of a story;
• a story you wrote or moved to Ready failed, was parked or was merged (not when you pressed Merge yourself).

If you are a member of a team but did not start the run, you see the story’s line in the team’s chat without being notified. Questions that someone else answered say who answered, in the chat and in the story’s history (“Spec fixed as Paula asked”).

Tell kanman a message did not help

When kanman writes to you on its own (a run is ready for review, a report is in, a story was merged), the message has a thumbs-down button, Not helpful, at its top right: point at the message to see it, on a phone it is always there. Press it when a message told you nothing you needed. A short line under the message confirms it, with Undo. Only the mark is counted, never the message, to measure how useful kanman’s own messages are; nobody else sees your mark.

Ask kanman

Type into the box at the bottom and press Enter (on a phone, tap the send button). Shift+Enter starts a new line. kanman shows that it is thinking the moment you send, also on a slow connection. While kanman answers, its reply appears word by word; Stop ends the answer early.

Without a connection (on a train, while your VPN reconnects) a line above the box says that you are offline. You can keep writing: your message waits in the conversation (“Sends when you’re back online”) and goes out by itself when the connection is back. The conversation then catches up on what happened in the meantime, including decisions you answered on another device.

The chip on the left of the box says which chat you are in: Personal or a team. On a phone, the mind button (a brain icon labelled, for example, Open the mind of Sandbox Team) sits right next to it; the team’s settings are the Team tab beside the Mind. In a team’s chat, kanman answers about that team. In a chat without messages yet, suggestions above the box help you start; once the chat is under way, they make room for the messages. You can always ask the same in your own words:

A question in Personal that is not about one team (“What should we watch out for when we roll out the export?”) is answered right there. Work you describe in Personal still goes to a team: kanman asks which one when your message does not say it.

Press N anywhere in the app to jump to the box of the chat you are in.

Describe new work

Describe work in the team’s chat: paste a requirement, a bug report, an e-mail or a few sentences into the box and send it. A long paste stays one message and folds to its first lines (Show all opens it), so kanman’s drafts stay in view, also on a phone. kanman drafts stories and shows them as a card, one row per story:

1. To write every ready draft, click Write all 2 ready stories; Reject the ones you do not want first. To pick single drafts, Approve them: the button then says Write 2 stories (the number follows your choice).
2. kanman writes them to your tracker and the card says “2 stories written”.
3. To edit acceptance criteria first, click Open review. The review page shows each draft’s open questions at the top, then every criterion and the test plan; Back to the conversation brings you back to the card.

kanman drafts with what it knows about the team’s repository: the code host (GitHub or GitLab), the default branch, the package manager and lockfile, the frameworks and scripts in package.json, the CI system, the top-level folders and the start of the README, plus passages from the team’s documents if you added sources under Mind > Memory > Sources. So drafts use your package manager’s commands and propose changes for your CI (a .gitlab-ci.yml on GitLab, never GitHub Actions) instead of asking about them. If your workspace runs everything on its own runners, the runner reads these facts itself; your code never passes through kanman’s cloud.

When a request is too vague to draft, kanman asks first (“Before I draft, I need to know: …”), usually with what it would choose. Just answer in the chat: kanman adds your answer to the same request and drafts the stories from both. Navigation and status questions you send in between keep their meaning.

Answer kanman’s questions

kanman settles design and code details on its own and writes them down as assumptions on the story. When a draft still needs something only you can decide, kanman says so right after drafting, for example “I need 2 answers from you before SBX-301 is ready.”, and the questions sit under their story in the card, numbered. Each one says whether it Blocks approval. For each question you can:

• click Answer, type a short answer and press Enter: kanman works it into the acceptance criteria and the test plan, and the question disappears,
• click Let kanman decide: kanman takes its own pick and notes it as an assumption,
• click Not relevant: the question is dropped.

Or just reply in the chat, by number: “1: only the section navigation, 2: you decide, 3: not relevant”. kanman maps your answers to the questions, updates the draft and tells you whether the story is ready now or how many questions still block it (“I worked your answers into SBX-301. It is ready to approve.”). A free-text reply without numbers works too; if kanman cannot tell which question it answers, it asks you to number your answers. A message that is not about the questions, such as new work, is handled as usual.

Clicking Approve on a draft that a question still blocks does not approve it: kanman says why and moves you to the first blocking question. Workspace owners and admins can choose Approve with open questions instead; the questions then stay on the story as notes. See Assumptions and open questions.

Long requirements are always treated as work to draft, even when they mention words like “status”, “overview” or “priority” (for example a UI label you quote). Short questions and commands keep their meaning.

You can attach text files (Markdown, plain text, CSV, JSON, YAML or logs), PDF files and Word documents (.docx) with the paperclip or by dropping them on the box; images work the same way (see Attaching and pasting images). Their text is sent with your message, and the conversation shows each file by its name; kanman drafts from the file like from a pasted requirement.

PDF and Word files. A requirements document, a specification or meeting notes can go in as they are. While kanman reads the file, it shows as “Reading spec.pdf…” in the message box, and Send waits until its text is there. A file can be up to 10 MB, and its text has to fit into one message (about 20,000 characters, together with what you type); for a longer document, attach the part that matters. The file stays on your message: click its name to open it again. A scanned PDF has no text kanman can read; paste a screenshot of the pages instead. Older Word files (.doc) are not supported: save them as .docx first. Reading the file works the same when your workspace runs on its own runners, because no model is involved; the file is stored with the chat, like your images.

Refine an existing story or issue. A vague issue that is already on the board or in your tracker can go through the same drafting: type “refine PU-43” (or “refine #12” for a GitHub or GitLab issue), or open the story and click Refine with kanman, which puts that message into the team’s chat for you. kanman reads the issue, drafts from it like from a pasted requirement, and asks what is unclear. When you write the first draft, kanman updates the existing issue and its story in place, so the key, links and history stay; further drafts become new stories. Stories that are already being worked on, closed stories and drafts that are still waiting for review are not refined.

This is how it looks: a requirement for the Sandbox Team goes in in Personal, continues in the team’s chat, kanman drafts three stories there, and approving them writes them to the tracker.

Attaching and pasting images

Show kanman what you mean: a screenshot of a bug, a sketch of a screen or a photo of a whiteboard. There are three ways to add an image to your message:

• Paste it. Take a screenshot or copy an image, click into the message box and press Ctrl+V (on a Mac Cmd+V).
• Drop it. Drag one or more images onto the message box.
• Pick it. Click the paperclip next to the send button. On a phone it opens your photo library or the camera.

You can attach PNG, JPEG, WebP and GIF images, up to 6 per message and up to 10 MB each. Very large images are scaled down before they are sent, so they go out quickly and still show every detail kanman needs. Each image appears as a small preview in the message box with a button to remove it; you can send images with or without text. Text files (Markdown, plain text, CSV, JSON, YAML and logs) work as before.

Your sent images appear in your message. Click one to see it larger; Escape or Close brings you back.

kanman looks at your images when it answers. Describe a bug and paste the screenshot, and kanman uses what it shows as context when it drafts the stories, for example the screen, the visible error and the steps that lead to it. If your team’s model cannot read images, kanman says so in one sentence and answers from your text.

Privacy. Images are stored with the chat they were sent in: in Personal only you can see them, in a team’s chat everyone on the team can. They are shown to you through links that expire after a short time. If your workspace runs kanman’s work on its own runners, your runner fetches the image through a short-lived link and sends it to your own model, never to kanman’s cloud models.

Slack and Microsoft Teams. kanman looks at images in the app only. When you send kanman a direct message with images, the copy of that message in your app conversation notes that the images stayed in Slack or Microsoft Teams. A message with only an image gets a short reply that points you to the app.

When a Personal conversation turns to one team

Conversations often start in Personal and then become work for one team, for example when you write down what a team should build. kanman notices and asks first, it never moves a conversation on its own: a card says Continue in Sandbox Team and quotes your request.

• Continue in Sandbox Team opens the team’s chat. It starts with a short line “Started in Personal” and your request, and kanman carries on there, for example by drafting the stories. In Personal the card folds to “Moved” and a short line stays behind: “Continued in Sandbox Team”, with a link that takes you straight to the team’s chat. In the team’s chat, Back to Personal takes you to where it started.
• Stay here keeps the conversation in Personal.

You can also move a message yourself: point at one of your messages (on a phone it is always shown), open its menu ⋯ and choose Move to Sandbox Team. The team’s chat opens with your message in the box, ready to send or edit.

When your workspace runs on your own runners

If your workspace runs everything on its own runners, kanman’s answers that need a model come from your runner. When no runner is online, kanman says so in the message (“Your runners are offline, so I can’t answer yet”) and answers in the same message as soon as a runner picks it up. A small note under the answer names the runner it came from, for example “via runner-eu-1”. Status answers, explanations from run records and everything kanman tells you on its own never wait for a runner.

Link to a message

Every message has its own address in your workspace. Links from Slack or Microsoft Teams (“Open in kanman”) and search results open the chat the message belongs to (Personal or a team’s chat), scroll to it and highlight it briefly.