How Kimtu works

Kimtu holds your architecture, documentation and system context in one place, for developers and for AI agents. Today that context gets in through one committed configuration file, generated with the help of your coding agent.

Early MVPThis is the first integration approach. We are building toward making architecture and product knowledge available to your development tools and AI agents without this manual import step.

The pipeline today

Kimtu never reads your source code. A coding agent running on your machine analyses the repository, writes a kimtu.yaml describing the architecture, and uploads that single file. Everything in the workspace is rendered from it.

01
The agent analyses the repositoryIt identifies services, datastores, queues and external providers, and the calls and events between them. It runs on your machine, in the coding agent you already use.
02
It writes kimtu.yaml at the repository rootCommit the file. From then on it is reviewed like any other config, in pull requests.
03
Kimtu builds the mapEach component becomes a page; each call, event and datastore becomes an edge on the architecture graph and a row on the data flow view.
04
Markdown is rendered alongside itIf you opt in, the markdown under the documentation directory renders as pages in Kimtu's own styling — with the file's path, and its commit and date when the repository is a checkout.

What Kimtu receives

SentThe kimtu.yaml file, and — only when you tick the box — the markdown under the documentation directory it declares, plus mkdocs.yml. Nothing else.

Never sentSource code, environment files, lockfiles, secrets, credentials or build artefacts. Kimtu runs no model of its own over your repository.

Where this is going

Kimtu’s purpose is not to import an architecture once. It is to hold the architecture and product knowledge of a system so that developers can understand how it fits together and AI agents can build against it with the same context a senior engineer has.

Available now: AI-assisted architecture definition: your coding agent generates the configuration, Kimtu renders the map, the data flow and the documentation.
In progress: Continuous integration with the development workflow, so the map updates as the system changes rather than on a manual import.
Planned: Direct agent access to the knowledge layer, so an agent working in your repository can query architecture and product context as it works.

Where to go next

kimtu.yaml referenceEvery field, with its type Component typesWhat each type renders as Documentation filesMarkdown and mkdocs.yml