Skip to main content

Quick Start

Runnable first steps with kapi — no install, no setup. These are ad-hoc runs: one command, one file, configured by flags. It's the fastest way to see what kapi does, and the right shape for one-off jobs, scripting, and CI. For content you localize repeatedly, you'll graduate to a project — see the end of this page.

Most commands below carry a ▸ Run button. Pressing it runs the real kapi binary in your browser — no install, and nothing leaves your machine. The first run takes a moment to start; after that it's instant. The commands that need an API key are shown as static blocks; run those after installing and adding a credential.

Pseudo-translate a file​

Generate pseudo-translations to test your UI for truncation and layout issues. Press ▸ Run — it seeds a small messages.json and writes messages-qps.json, which you can preview in the Files panel:

kapi pseudo-translate messages.json -o messages-qps.json

The output is accented text like ▒ Ĥéļļö, Ŵöŕļđ! ▒ that exposes untranslated strings and layout problems. With no --target-lang, kapi uses the qps pseudo-locale; pass --target-lang to target a real locale.

Count words​

Estimate translation volume for cost planning:

kapi word-count messages.json

Explore formats, tools, and flows​

List every format kapi understands — no input file required:

kapi formats

List the available processing tools, composable flows, and presets:

kapi tools
kapi flows
kapi presets list

Edit before you run​

With editable, the command is placed at the prompt instead of running automatically — press Enter when you're ready, or tweak it first:

kapi word-count messages.json --json

For a full interactive terminal, open the CLI Playground.

Translate with AI​

Translate a file with an LLM, then run a QA check in the same flow:

kapi run ai-translate-qa -i input.html -o output.html --source-lang en --target-lang fr

This needs a provider credential, so install first and add one with kapi credentials add — see Installation. When a brand voice profile is bound on the flow, its guide is injected into the translation prompt, so output is on-brand at generation time.

Keep AI output on-brand​

Print a brand voice guide to paste into your AI assistant's context, starting from a built-in starter pack:

kapi brand guide --pack friendly-dtc

Score a draft against a profile. --min-score makes it a CI gate — kapi exits non-zero when the score is below the threshold:

kapi brand check --profile-file brand.yaml --min-score 80 release-notes.md

Input can be a file argument, --input-text "…", or piped via stdin. Rewrite off-voice content — forbidden terms, competitor mentions, jargon:

kapi brand rewrite --profile-file brand.yaml --input-text "Leverage our solution to drive synergy."

Five starter packs ship built in: professional-b2b, friendly-dtc, technical-docs, marketing-blog, and customer-support. See Brand Voice.

Manage terminology​

# Import terms from CSV
kapi termbase import terms.csv --format csv -s en -t fr

# Look up terms in text
kapi termbase lookup "authentication module" -s en -t fr

Use it inside your AI assistant​

Expose brand and terminology tools over MCP so your assistant stays on-brand while it writes:

kapi mcp

See Using kapi with Claude for the skill-based workflow, or Using Kapi with AI Assistants for MCP setup with Claude Code, Cursor, and Windsurf.

Next: set up a project​

The runs above are ad-hoc — perfect for a quick check or a CI step. Once you're localizing the same repository or content set more than once, capture the languages, content patterns, and flows in a .kapi project so you stop repeating flags and start accumulating translation memory.

Next steps​