Crowdin Skills

Crowdin Skills are instruction packages that teach an AI coding agent how to work with Crowdin. Once they are installed, the agent knows the Crowdin CLI commands, the configuration file format, the API client, and the query languages, so you can describe a localization task in plain language instead of looking up the syntax yourself.

Skills do not connect the agent to your Crowdin account. They work inside your development environment, next to your source code, unlike the Crowdin MCP Server.

Available Skills

• crowdin-cli – Syncing source files and translations between your project and Crowdin with Crowdin CLI v5.
• github-action – Setting up and debugging the crowdin/github-action step in a GitHub workflow.
• create-app – Scaffolding, developing, and publishing a Crowdin app.
• context-extraction – Writing context for the source strings that need it.
• crowdin-context-cli – Working with the crowdin context commands and the JSONL format they use.
• crowdin-api-client – Using the JavaScript and TypeScript API client, including pagination, uploads, and error handling.
• croql – Writing and optimizing CroQL expressions for strings, translations, TM segments, and glossary terms.
• graphql – Writing and debugging Crowdin GraphQL queries against the actual schema.

Installation

Skills work with Claude Code, Cursor, OpenAI Codex, Gemini CLI, GitHub Copilot, OpenCode, Cline, Windsurf, and other AI coding agents. Each client adds them in its own way, so follow the installation instructions in the repository.

Install Crowdin Skills

Enriching String Context

Context tells translators and AI models how a string is used: where it appears in the interface, what the placeholders hold, whether a short word is a button or a page heading. With Skills installed, the agent writes that description from the code where the string is used.

Two skills cover this flow: crowdin-context-cli for the commands and the file format, and context-extraction for what a useful context description looks like.

1. The agent runs crowdin context download and receives a JSONL file with your source strings.
2. The agent traces where each string is used in your codebase and writes a short description of its purpose and placement.
3. The agent runs crowdin context upload, and the context appears in your Crowdin project.

The download command takes filters, so the agent can work on a single branch, a set of labels, or only the strings that have no context yet. Run crowdin context status to see how much of your project is covered, and crowdin context reset to remove the context generated by the agent while keeping the context written by hand.

Read more about the crowdin context commands.

Skills and the MCP Server

Skills and the Crowdin MCP Server both help an AI agent work with Crowdin, in different places.

• Crowdin Skills teach the agent how Crowdin’s developer tools work. The agent applies that knowledge locally, against your source code and your crowdin.yml.
• Crowdin MCP Server connects the agent to your Crowdin account, so it can read and change project data directly.

You can use both: Skills to run the right command in your repository, and the MCP Server to check the result in your project.

Read more about the Crowdin MCP Server.