Context is one of the most effective ways to improve translation accuracy. Without it, translators and AI models are left guessing the meaning of short strings, UI labels, and ambiguous phrases – leading to translation errors and prolonged review cycles.

Instead of using isolated scripts or manually configuring individual AI provider API keys, you can now harvest and enrich string context directly within your development environment using Crowdin CLI and Crowdin Skills.

Key benefits

• High context precision: AI agents analyze your actual codebase to generate concise 1-3 sentence descriptions detailing UI element types and exact placements.
• Native agent integration: Works with Claude Code, Cursor, Copilot, Gemini CLI, and other Agent Skills-compatible tools.
• Granular filtering: Target only the strings that actually need context (for example, missing descriptions, specific branches, or recently added labels).
• Safe and non-destructive: Download, review, and upload context on your terms, with dedicated commands to reset AI-generated context without touching manual entries.

Workflow

Step 0: Install Crowdin Skills

Install the Crowdin Skills to teach your AI coding agent how to extract context and execute Crowdin CLI commands:

npx skills add crowdin/skills

(You can also install via Claude Code plugin, Gemini CLI extensions, or GitHub CLI).

This gives your AI agent two key skills:

• context-extraction: Teaches the agent best practices for writing string context (focusing on ambiguous short words, plurals, UI element type, placement, and safe editing rules).
• crowdin-context-cli: Teaches the agent how to manage context files and execute CLI download and upload commands automatically.

Step 1: Download strings context

Export your project's strings to a local .jsonl file ready for AI enrichment:

crowdin context download

By default, this creates a crowdin-context.jsonl file. You can apply filters to focus only on strings that require attention:

Step 2: Enrich context with your AI agent

Your AI agent already understands your local codebase. Once crowdin-context.jsonl is present, prompt your agent:

Using crowdin skills, enrich the context for empty strings in crowdin-context.jsonl

The agent will locate where each string is used in your code and fill the ai_context fields with accurate, human-readable descriptions.

Step 3: Upload enriched context

Push the updated context back to Crowdin:

crowdin context upload

(If you used a custom path during download via --to, use --from during upload).

Utility commands

Supported AI environments

Crowdin Skills are compatible with:

• Claude Code (/plugin install crowdin@crowdin-skills)
• Cursor / Windsurf / Cline / OpenCode (npx skills add crowdin/skills)
• Gemini CLI (gemini extensions install https://github.com/crowdin/skills)
• GitHub CLI / Copilot (gh skill install crowdin/skills --all)