About Wakagi

Wakagi is where you and your AI design a project, split it into tasks, and build it together. It is made for one person and one AI, not for a team.

A chat is gone when the window closes, and every new prompt starts from nothing. Wakagi gives the two of you a place to talk about the work, task by task, and keeps what was said where it belongs: on the task, for the next session or the next AI to read. Working together becomes simple. The AI asks for what is waiting, gets each task with what was said since it last spoke, does the work, and hands it back where you will look for it. It costs you less: nothing is read twice, nothing is pasted into a prompt, and the tokens go into the work rather than into repeating yourself. And it stays safe: the AI cannot start what you have not made ready, cannot declare its own work finished, and can do no more than its link allows. Wakagi is not a task manager with an AI feature. It is how you and your AI talk about a project.

Wakagi (若木) is Japanese for a young tree. Every project starts as one, and grows as you work on it.

The talk is the task

Most of the time on a task goes into talking about it, not into typing the code, and elsewhere that talk is lost in a chat somewhere else. Wakagi keeps it on the task it belongs to.

While you work with your AI in Claude Code, you say "start conversation #14". From then on, what you type and what the AI answers is kept on task 14, until you say "end conversation". Nothing is sent while no conversation is open.

That conversation is the memory of the project. A new session, or another AI, reads why a decision was made, not only what it was. Whoever is free thinks the work through, and whoever does it picks it up from the task, without either of you repeating it.

The AI's name in a conversation is the model that answered. It cannot choose its own.

The AI is a user, not a plugin

Elsewhere an AI is bolted on: a button that summarises a ticket, a bot that writes a comment. In Wakagi the AI is the second user of your project. You give it a link, and the link comes with a manual that tells it how the project works and what it may do. It reads tasks and changes them the way you do.

One call tells it what waits for it: the tasks you asked it to look at, with what was said since it last spoke, and the tasks that are ready to be worked on. Nothing is read twice, nothing is missed, and you never paste a ticket into a prompt.

You decide what each link allows: only seeing the tree, reading the tasks, or changing them. You can take a link back at any moment. A link that only shows the tree is also how you show your project to someone else.

One person, one AI

Task managers were made for teams: people assign work to each other, report on it, and look at boards to see who is behind. You are one person with one AI. Nobody assigns you anything, and there is nobody to report to. What you need is to think a project through, keep what you decided, and let your AI find the work without being told twice. So there are no sprints, no boards, no story points, no roles, and no workflows to configure. There is a tree, a conversation on every branch, and five states.

A tree, not a list

Elsewhere a task is a card in a column or a row in a backlog, and a big project is a thousand of them. In Wakagi the trunk is the project itself. Every other task is a branch, and it grew from another one: a task is never added to the end of a list, it is split off the task it belongs to. A big idea becomes a few parts, each part becomes smaller pieces, and the smallest ones are the work you actually do.

So the shape of the tree is the shape of your project. A task that was split has no state of its own: it is as far along as the tasks under it, and finished when they all are. You never tidy up a parent, and you never close an epic by hand.

Every task has a short number, the one you say and type: "task 14". A number is never given twice, not even after the task is gone, so a number you wrote down somewhere always means the same task.

You look at it, you do not read it

Instead of a dashboard, the project is a cherry tree, standing in a Japanese garden, with the names of the tasks carved in the bark. Walk around it, up and down the trunk. A task that is not ready is a bud, one being worked on is opening, a finished one is in full bloom. A bug is a caterpillar on its branch; once fixed, it is a butterfly. A task waiting for an answer has a bird singing on it. Where the project is growing and where it is done is a glance, not a report.

The list is there too: every task in its place in the outline, with its conversation, its plan and its files. You can see everything, only what is not ready yet, what the AI has given back, the bugs, or the tasks the AI has answered you on.

Five states and two rules

No workflow editor. A task is in one of five states:

not ready

Where every task starts, and where you put one back when you are not happy with it.

ready

You are happy with the conversation and the plan. It may be worked on.

working

The AI has been told to start, and has.

work done

The AI has finished and given it back, for you to check.

finished

You are happy with the result.

Three of these are yours: not ready, ready and finished. Two are the AI's: working, which it may only set on a task that is ready, and work done. It cannot start what you have not made ready, and it cannot call its own work finished. That is the whole process, and it is built into Wakagi, not left to the AI's good manners.

Projects grow, and so does the garden

Projects grow in directions nobody planned, and elsewhere moving work between them is a chore. Here, when a branch has grown big enough to be a project of its own, plant it: the branch, with everything that grew from it and all its conversations, becomes a new tree. A graft moves a branch onto another tree you already have. The moved tasks get new numbers in their new tree, and their old numbers say where they went, so nothing you wrote down before gets lost.

Projects that belong together can be put in a grove.

Yours, on your side

Your browser keeps a copy of your projects, so they open at once, and a connection that drops does not stop you. What you change goes to Wakagi as soon as it can.

You sign in with your email address. Wakagi sends you a code, you type it, and that is all: no passwords to invent, forget, or have stolen.

Wakagi keeps your email address, to sign you in, and your projects: the tasks, their plans, the names of your files, and the conversations you opened. Never the content of your files. Never your sessions with your AI outside an open conversation. Nothing is sold or shown to anyone else.

Alpha

Wakagi is in a closed alpha: it is used by a few people who were invited, and you can sign in only with an address that was invited. If you would like to be one of them, write to admin@wakagi.dev. Anyone can try it without an account: the demo is a whole project, in your browser, yours to change.

Price

The first 14 days are free, with everything in them and no card needed.

After that Wakagi costs €10 a month in Europe, $10 in the United States and £9 in the United Kingdom. The checkout shows the exact amount, taxes included. Stop whenever you like: the month you paid for stays yours.

If you don't subscribe, nothing is lost. Your projects stay there to read, and everything opens again the day you subscribe.

Made in Wakagi

Wakagi is made by one person and one AI, the way it is meant to be used. It was designed in its own conversations, and it grows the way your projects will: one branch at a time.