Skip to main content

Kapi

Kapi reads, translates, and ships content in any format — faithfully — and verifies it with checks (terminology, QA, brand voice) that act like tests for AI output. It is built on the open-source neokapi framework and comes in two forms that share the same engine, tools, flows, and .kapi projects:

  • Kapi CLI — a single binary that operates directly on files. Best for automation, scripting, CI gates, and offline work; no project or server required. It also serves brand and terminology tools to your AI assistant over MCP.
  • Kapi Desktop — a visual companion app for building flows, running tools, managing plugins, and storing AI credentials — without writing YAML or remembering flags.

Because both front-ends run the same engine, a workflow built in one works in the other: the flows and the project file are identical.

Extract content from a file, translate it, run checks, and write it back in the same format — a full bilingual round-trip from the terminal:

Command-line interface​

Run tools and flows straight from the terminal — kapi pseudo-translate, kapi ai-translate, kapi brand check, kapi run <flow>. See the Kapi CLI overview for the command surface and the task-by-task recipes you can run in the browser.

Desktop app​

Prefer a visual interface? Kapi Desktop wraps the same tools and flows in a native app with a flow editor, live runner, plugin manager, and credential vault. The Kapi Desktop overview walks through your first project.

Projects and ad-hoc​

There are two ways to drive Kapi:

  • Project — the recommended day-to-day model. Capture the workflow once in a portable .kapi recipe (languages, content patterns, flows, plugin requirements, defaults) and replay it with kapi run <flow>. The project store accumulates translation memory as you work, and the recipe is shareable via git. Best for any repository or content set you localize repeatedly — create your first project.
  • Ad-hoc — run a tool or flow directly on files, configured by flags (or, in the desktop app, clicks): kapi ai-translate -i file.json --target-lang fr. Nothing is saved — the quick path for one-off jobs, scripting, and CI.

Kapi finds your project automatically. Like git, it walks up the directory tree from wherever you run it; if it finds a .kapi recipe, it runs in project mode with that project's languages, defaults, and flows. Otherwise it runs ad-hoc.

Both are the same commands over a different binding (where content enters and leaves): ad-hoc writes files, while in a project a run commits to the store and kapi merge materializes the localized files. See Modes & bindings for the full model — the source/sink binding diagram, the two project loops, the portable .klz parcel, and when it makes sense to mix.

See the project file reference for the full .kapi format.

Install​

See Installation for the CLI (Homebrew or download) and Kapi Desktop (macOS cask).