Docs

Docs your team writes. Your agents read.

Docs are the pages your team writes together and your coding agents read and edit: runbooks, architecture notes, product briefs, whatever your project needs written down. They're the written part of the context layer: one set of pages that people and agents share.

Pages in a tree

A page is one document of your docs, written in Markdown and nested under other pages: a knowledge base for AI agents and for the people they work with.

Coding agents read Markdown easily and efficiently, so a page costs them few tokens, and people read it just as well. Your team writes in the web app and your agents from their sessions, on the same pages.

Drafts, versions and publishing

Publishing decides what your agents build from: a page stays a draft while you write it, and your agents work from it once you publish.

Work in progress never steers an agent, and a run under way isn't thrown off by an edit you publish halfway through. Every version is kept to compare and go back to.

Diagrams in the page

A diagram lives on the page it explains: Mermaid written as text and drawn in place, or a drawing or image from your library.

A Mermaid diagram is text, so your agents can draw one for you and update it when the design changes, and every change shows in the page's versions.

Comments where the text is

A comment thread is a conversation on one passage of a page, kept beside the text, where your team and your agents settle a question together.

A teammate marks a section with a question, and an agent answers beside the text: the assistant in the web app, or the coding agent in your IDE after a look at your code. An agent working from the page sees open questions too, and asks instead of building past them.

Linked to the work

Link a page to the work it explains, such as a board item, a record, a workflow or another page, and every link shows its current title.

Point an item at the one section it builds, so its agent reads that part, not the whole page. Rename anything, and no link breaks.

Read by the right task

Every task starts with the docs it needs: a workflow hands each node its pages, and the context step adds the ones that apply.

Your agents spend their effort on the work, not on hunting for the right page.

A product brief your implementation workflow builds from

You write the brief for annual billing. Your implementation workflow builds from it.

  1. Write the brief

    You start a draft under Product, with sections for the goal, who it's for, what done means and what's out of scope. A teammate asks on the pricing passage whether checkout shows yearly prices, and your coding agent answers in the thread.

  2. Publish it

    Once the question is settled, you publish. The board item “Add annual billing” links to the brief's “What done means” section.

  3. The workflow reads it

    When the item reaches its column, your implementation workflow starts, and its Plan node reads the published brief.

  4. The agent asks where the brief is silent

    The brief says nothing about refunds on annual plans. The agent leaves a comment on that passage instead of guessing, and you answer it there.

You add the answer to the brief and publish again. The change is built from the brief, and every later run reads the same answer.

Questions about docs

Are docs a knowledge base for AI agents, or for people?

Both. Pages are Markdown, which agents read easily and efficiently and people read just as well. Your team works on them in the web app and your coding agents from their sessions, on the same pages.

Can my agents edit docs?

Yes. Your agents write and update pages like any teammate, and every change becomes a version you can compare and go back from, with the agent named in the history.

Do agents read my drafts?

No. Your workflows read published pages only, so your agents build from what you've published. An agent that comes across a draft knows it's one.

Which diagrams can a page hold?

Mermaid diagrams, written as text and drawn in place, and drawings or images kept in your library.

Can agents read and answer comments?

Yes. Ask a question on a passage, and an agent answers in the thread: the assistant in the web app, or the coding agent in your IDE. An agent working from the page also sees open questions where they apply.

Can I export my docs as Markdown?

Yes. Every page is Markdown. Export everything as plain files at any time.

Write your docs where your agents read them

Start a page in the web app, or have your coding agent write the first draft.