Skip to main content
neokapi

neokapi — The faithful, format-aware content engine for people and AI agents

An open-source engine in Go that parses localization, document, and data formats into a faithful content model — then translates it, leverages memory, and runs checks for terminology, QA, and brand voice, whether a person or an agent wrote it.

Try Kapi with ClaudeGet Started

Read. Parse any of 50+ formats into one faithful content model.

neokapi is a re-imagination of the Okapi Framework in Go, built for humans and agents.

One engine, many applications

neokapi is a Go framework. Use it directly as a library, or through kapi — the CLI and desktop app built on it.

The engine

Go framework

Format-aware readers and writers, a faithful content model, and a streaming pipeline of composable tools — embed it directly in your own Go programs.

Go quickstart →
Built on it

kapi — CLI & desktop

Drive the engine from the command line or a visual desktop app: extract, translate, run checks, and manage .kapi projects — no code required.

Use kapi →
In the familykapi-react — a zero-config i18n library for React. Its own framework, powered by neokapi under the hood for build-time string extraction and catalog compilation.In the familyCLI Tools — kgrep, ksed, kcat: format-aware grep, sed and cat that operate on the translatable text inside .docx, JSON, XLIFF and more.

Formats & plugins

Readers and writers for localization, document, data, subtitle, and office formats, extended by plugins and a bridge to the Java Okapi filters.

Formats →

AI-native tools

LLM-assisted translation, QA, terminology, and review compose in the same pipeline as machine-translation backends and rule-based checks.

AI translation →

Streaming pipeline

Tools run in parallel and stream results as each part is ready, so large files and many languages process fast.

Architecture →

Interchange with any TMS

Extract bilingual XLIFF 2.x or PO for Trados, memoQ, Phrase, or Crowdin, merge the translation back through a faithful skeleton, and keep translation memory and terminology in the loop.

Interchange →

Project model

Capture languages, content patterns, and flows once in a committed .kapi recipe; run flows with no repeated flags. Translation memory accumulates, and git-style discovery finds the project from any subdirectory.

Create a project →