Skip to main content

How Bowrain and kapi fit together

Kapi is the Apache-2.0 toolchain Bowrain is built on: it reads and writes content formats, runs checks, drafts and adapts text, and works entirely from local files with no server or account. Bowrain is the governed platform — shared voice, vocabulary, and content memory, collaborative editing, review, automation, connectors, and version history.

kapi appears in two roles here, and they are worth keeping apart:

  • The engine underneath. The same format handling, checks, and flow execution run inside Bowrain's server. This is invisible in day-to-day use; it is why the platform behaves identically whether content arrived from a content platform or from a repository.
  • One connector into the platform. With the bowrain plugin installed, kapi connects a developer's checkout to a workspace — the developer and CI route described in the kapi connector. It sits alongside the content-platform, design, and repository connectors rather than in front of them.

The boundary

kapi owns the local files and the project configuration. The kapi.yaml recipe — content collections, flows, plugins, languages, brand, and the server: block — is authored and versioned in the repository with everything else. Bowrain never writes it.

Bowrain's local footprint is cache and speed only, never a source of truth. The Bowrain desktop app is a working copy of the server: a content cache, an offline edit queue, and memory and vocabulary mirrors. It does not author local files or source projects from a filesystem — sourcing from a filesystem or a git checkout happens server-side through connectors, on the host the server runs on.

At a glance

kapiBowrain
ShapeA CLI + desktop app you installA server + web and desktop clients
UsersOne person, one checkoutOne workspace — you, and your team when you have one
StateLocal files + local memory and termsHosted, versioned content store (the desktop holds a cache only)
Brand & vocabularyA profile you carry in filesShared, governed, and learned from corrections
Content sourcesLocal files you ownEvery connector — content platforms, design tools, repositories, checkouts
AutomationLocal recipe hooksServer-side, event-driven
CostFree, open sourceHosted plans / self-host

kapi on its own is enough when

  • One person works from one repository they own.
  • The work is drafting, checking, or pseudo-translating files from a terminal or a desktop app — no account, offline by default.
  • Checks and language work are being wired into CI, or into an AI assistant over MCP.

Reach for Bowrain when

  • Content lives in systems beyond one checkout — a content platform, a design tool, a repository nobody wants a pipeline in — and should sync through connectors.
  • Several projects or several surfaces should draw on one memory, vocabulary, and voice that compound across all of them, kept current on the server rather than only when someone runs a command.
  • Several people — writers, reviewers, editors — work on the same content and need to see each other and edit together.
  • One brand voice and vocabulary should be shared and governed across everyone and every AI tool, with history and audit.
  • Corrections should compound: a fix made once becomes a versioned, enforced check.

The two are not alternatives, and neither is a stage you graduate from. See the introduction for what the platform does, and Connectors for every route into it.