Build with Translatize
Translation management with git-like branching. These docs cover the features your team uses in the app and the tools that connect Translatize to your code: the CLI, the CI templates, the MCP server for AI agents and the integration API they share.
Features
Learns from approved and edited translations, never from AI output, imports or merges; suggests them in the editor on every plan and fills exact matches unbilled when AI translation runs.
Translates with your key names, descriptions, character limits, glossary and style guide as context, on Professional and Agency. Labels marked "Do not translate" stay the same in every language, on every plan.
Ten rule checks on every plan, a merge gate that warns or blocks, and AI review on Professional and Agency.
Show translators where a string appears: paste or upload in the editor on any plan, or capture from your Playwright tests on Starter and above.
Publish translations to our CDN so text fixes reach your app without a new release, with release history and rollback. An add-on from Starter up.
A Translatize branch and a "Translatize" check for every pull request that changes your source strings. Professional and Agency.
Who changed what in a project or on your account, kept for 30 to 365 days depending on the plan, with CSV export of the project log on Professional and Agency.
Tooling
Pull, push and compare translations, run quality checks, upload screenshots and publish releases from your terminal or CI.
Run the CLI in GitHub workflows – push keys on merge, pull translations, and fail pull requests on missing strings or QA errors.
A CI/CD Catalog component and copy-paste jobs that run the CLI inside your GitLab pipelines.
Let AI agents like Claude Code and Cursor manage your translations over the Model Context Protocol.
The token-authenticated REST API every tool is built on – labels, branches, exports, AI translation, quality checks, screenshots and releases.
How it fits together
Every tool speaks to the same integration API. The CLI is the workhorse; CI runs the CLI; AI agents reach the same endpoints through the MCP server.
@translatize scope, all at version 0.2.0:- @translatize/cli – the
translatizecommand-line tool. - @translatize/core – the TypeScript SDK for the integration API.
- @translatize/mcp – the MCP server for AI agents.
- @translatize/ota – the over-the-air client your app uses to load published translations.
- @translatize/playwright – screenshot capture inside your Playwright tests.
npx @translatize/cli@latest.Where tokens come from
Every tool authenticates with an API token – a string that begins with mcni_, not your login. Create one in the app under Project Settings → Integrations. A token is bound to a single project and a base branch. Its value is shown once, so copy it when you create it.
Each token has a role: translator, developer or admin. You need the developer role or higher to create one, and a token can't have a higher role than yours. A translator token reads and writes translations; setting key context and tags, uploading screenshots and publishing releases need a developer token; merging a branch into main needs an admin token.
For AI agents, pick the Agent workflow branch scope (create-own in the API). Such a token can also create branches from its base branch, work on them and merge them back, but it can't reach any other branch in the project – so an agent can have a token of its own without access to the rest. The MCP Server and Integration API pages describe that workflow.