Kaiku

Task tracker · Wiki · Agents

An agent-native tracker and wiki.

Recorded inside the product, at the speed it actually moves. Sound on for the commentary.

In Kaiku, agents work in the tracker and the wiki the way the rest of the team does. Hand one an issue or call it in a comment, connect your own over MCP, and see every run on the issue it worked on. An agent proposes; a person approves. Tools written for the API most companies already run connect too.

30 days free. No card, no wallet, no sales call.

Three things it does differently

Agents are the point, not a plugin

The MCP servers your agents already use for the incumbent tracker and its wiki work against Kaiku as they are — the compatibility is on the wire, not in a bridge that lags a version behind. Kaiku also ships an MCP server of its own, and records every agent run against the issue it worked on: tokens by kind, machine time, who asked. You find out what the automation actually cost while you can still do something about it.

A task tracker for AI agents

A wiki that speaks the same protocol

Every project carries exactly one space, keyed like the project. You write Markdown; the wire carries the storage format the incumbent corporate wiki uses, so tools built for it read Kaiku without being told. Pages go out as Word documents, and the words inside an attached diagram are searchable — a drawing stops being a dead end.

The wiki beside the tasks

The data stays yours

One request gives you the whole project as a zip: structured JSON, readable Markdown, and every attachment. It keeps working after a subscription lapses — a service that locks your data behind a payment is a service nobody should pick.

Point your agent at it.

Kaiku speaks MCP two ways: through its own server, and through the ones your agents already have configured for the tracker and the wiki your company runs today. No account yet? Give your agent one address and ask it to sign you up.

New here: just the address, and ask your agent to sign you up

{
  "mcpServers": {
    "kaiku": {
      "type": "http",
      "url": "https://app.kaiku.tech/mcp"
    }
  }
}
claude mcp add \
  --transport http \
  kaiku https://app.kaiku.tech/mcp

No token and no form. The agent shows you the terms, asks for your email and a name for your company, and you read it the code from our email. It gets a token for the new workspace and reconnects with it.

Already have a workspace: its address and a token

{
  "mcpServers": {
    "kaiku": {
      "type": "http",
      "url": "https://<your-workspace>.kaiku.tech/mcp",
      "headers": { "Authorization": "Bearer <token>" }
    }
  }
}

Your workspace answers on its own address, and issues its own tokens from Settings. A read-only token cannot reach the MCP endpoint at all.

The servers you already run
The MCP servers built for the incumbent corporate tracker and its wiki work against Kaiku unchanged, because the compatibility is on the wire rather than in a bridge that lags a version behind. Nothing to install, nothing to fork.
And one of our own
Issues and their children, comments and the threads they hang in, wiki pages, attachments by path, project fields, and questions put to one person and waiting on an answer. An agent can read and write the same things a person can, and no more than the token allows.
What the automation cost
Every agent run is recorded against the issue it worked on: tokens by kind, machine time, who asked. You find out what a week of automation actually cost while there is still something to do about it.
A task tracker for AI agents

Bring the work you already have.

A company admin imports a project from the cloud version of a popular task manager: issues keep their keys, and their comments, files, custom fields, parents and links come with them. A check runs first and names what will not move — change history, sprints, logged time — before anything is written.

Kaiku Agents

It reads the issue and proposes. You decide.

Call one in a comment, the way you call a colleague — by name, in the thread where the question already is. It reads the issue, answers in the same thread, and says what it read to get there.

It never decides. A proposal is recorded as a question addressed to you, with no default answer, so the issue keeps who chose and what they chose — and silence chooses nothing.

Recorded inside the product, on a board made up for the camera. The answers are scripted; everything around them is the product. Sound on for the commentary.

Your access, never its own
An agent is a role of your workspace, not an account: no password, no token, no access of its own. It runs as the person who called it, with the permissions they have at that moment, and is clamped to the issue’s project — a question about one project never ships the other nine to a model. Its tools are reads only; the comment it answers with is written by Kaiku, not by the model.
What it read, and what it cost
Every answer carries the issues and pages its tools actually returned, so the reasoning can be checked against its sources. Every call is a row — who called which role, on which issue, what came of it and what it cost — so the bill for the automation is visible while you can still do something about it.
How agents work

Mini-CRM

Some teams keep clients, not a backlog.

Turn on the client book in a project and it reads as one: a row per client, the columns your business actually runs on, and the work for each client right under its row. Nothing leaves the tracker to get there — the board, search and your agents keep seeing the same issues.

Mini, because it is a table of clients with the work under each one — not a sales system with a deal pipeline, follow-up reminders or an email history.

Recorded inside the product, in a client book made up for the camera. Sound on for the commentary.

A row is a client
A client is an issue of its own type, so it comes with comments, files, history and watchers — and the same API and MCP as everything else. Open the row and the work on that client is underneath, filterable by tag.
Columns of your own
The project writes down the columns it needs, of six kinds: a list with colours, a country, a number, a date, a person, a line of text. A value that does not fit its column is refused on the spot, never quietly dropped.
Edited in the cell
The name, the owner and every column of your own change right in the table, instead of opening the card and coming back. It is the same write the card makes, so history, notifications and your agents all see it.
Out as a spreadsheet
One .xlsx with your columns in your order — the whole book, not the page on the screen. A number stays a number and a date stays a date, which a CSV opened in a spreadsheet never promises.
More about the mini-CRM

Case study · A repertory theatre

A season, run as project work.

A dozen productions share one board, and each opening night has thirty pieces of work in front of it. The calendar shows when each of them ran and what each is waiting on — and it draws a bar only between days that were actually recorded, never from a start somebody guessed.

Read the case

With a 90-second film from inside the product.

The calendar in Kaiku: a theatre season’s work as bars and points on one axis, today marked, and two issues with no dates named underneath.

What you get

The ordinary things a tracker owes you, and a few it usually does not.

Boards, sprints, backlog

Kanban and scrum boards, drag between columns, sprints that start and close. A board bigger than a page says so rather than quietly showing you a prefix.

Fields on a board, too

Your own fields are not only for clients. Any project can define them, search by them, colour its board cards by a list’s value, and mark a card with the field that matters.

Read it in your language

Issues, comments and wiki pages are translated automatically, on demand. The language picker searches by what each language calls itself.

Merge requests on the issue

Several GitLab repositories per project, because a tracker project rarely lives in one. A repository that did not answer says so instead of pretending there is nothing there.

Numbers that stay honest

Stock and flow kept apart — what is open now never quietly takes the date range that created-per-day takes. Archived issues are still counted, because a shelf is not a delete.

Archiving that is not deleting

Clear a board without losing anything, by hand or on a rule the project sets. Counted from when the task became done, not from when it was made.

Sign in without a password

Passkeys and Google sign-in, or a password if you prefer one. You cannot remove your last way in, which is the mistake that locks people out.

Notifications that say when

Mentions, watchers, reactions on comments, and a digest that respects what you asked for. Every timestamp keeps its exact value behind the friendly one.

Pricing

One plan for one person, and a team plan with our AI or with your own AI key. It costs what it costs, per person who uses it.

Trial

Freefor 30 days

  • Up to 2 people
  • 1 project
  • 1 GB of attachments
  • Every feature, none held back
  • No card and no wallet to start
Start free

Personal

$19.99per month

  • One person, one seat
  • The assistant without a ceiling, under fair use
  • Up to 100 projects
  • 20,000 issues and wiki pages
  • 10 GB of attachments
  • Bought in the iOS and Android app, or here on the web
Start free

Team

$7per person / month

  • Everyone you invite, billed per person
  • Up to 100 projects
  • 20,000 issues and wiki pages
  • 20 GB of attachments
  • +$20/month per extra 20,000, per workspace
  • +$10/month per extra 100 GB of attachments, per workspace
  • AI included — or your own key
  • $5.25 per person / month when paid for a year
Start free, pay later

Team, own AI key

$6per person / month

  • Everything in Team
  • No built-in AI: agents, the assistant and translation run on your company’s own Anthropic or DeepSeek key
  • The provider bills you directly; switched-on agents are a fixed fee each
  • Switch between the two team plans at any time
Start free, pay later

AI Pro, on Team

+$12.99per seat / month

For the people who lean on the assistant all day. A seat is given to a named person and moves when they leave; its holder’s assistant is counted on its own, under the same fair use as Personal, instead of from the team’s shared allowance. Agents stay per call for everyone. Not on the own-key plan, and no yearly discount.

Pay by card, as a subscription you can cancel at any time and refunded in full within 14 days, or in USDT or USDC, each period prepaid. Card orders are processed by our reseller Paddle.com, the Merchant of Record. Personal can also be bought through the App Store or Google Play in the mobile app.

Questions people actually ask

Which tools work with it, exactly?

Anything written against the REST API of the tracker most companies already run — the same request shapes, status codes and error formats, including the MCP server your agent uses for it today. You give the tool your workspace address and your token, and it does not learn it is talking to something else. The same holds for the wiki. Write to us and we will tell you plainly whether the tool you have in mind is covered.

Where does my workspace live?

On your own subdomain — yourcompany.kaiku.tech. Your project keys are yours: your CRM has nothing to do with anyone else’s, and issue keys never collide across companies.

What happens when I stop paying?

The workspace becomes read-only. Nothing is deleted, everything is still readable, and export keeps working so you can take your data with you. We keep it that way for 12 months and warn you well before anything is removed.

Can I pay by card?

Yes. A card pays as a subscription that renews each month or year until you cancel it, which you can do at any time; a card payment is refunded in full if you ask within 14 days. The card orders are processed by our reseller Paddle.com. You can also pay in USDT or USDC, each period up front, and you will always get a warning before one runs out.

What counts towards the 20,000?

An issue or a wiki page — one each, counted together across the whole workspace rather than per project. Archived issues still count: a shelf is not a delete, and a limit you can dodge by archiving is not a limit. Past the included amount it is $20 a month per extra 20,000, charged to the workspace and not per person — a team of three with a large archive pays the same for volume as a team of thirty.

What if I go over?

Nothing shuts off that day. You get 7 days to top up, and you keep writing the whole time. Going over is usually something you did not notice — an import, or an agent that filed three hundred issues overnight — and locking someone out of something they do not know about yet is a failure that looks like a fault.

What counts as a user?

A person with access to the workspace. Invite them and they count; remove them and they stop. Change the number mid-period and the difference is prorated rather than charged twice.

Can I self-host it?

Yes. Send a request for a self-hosted installation to hello@kaiku.tech and we will come back to you with what it involves.

Start with a month.

Two people, one project, everything switched on. If it does not fit, you have lost an afternoon and taken your data with you.

Create a workspace