The new Crowdin CLI context management commands make it easy to manage string context at scale, directly from your terminal.

In this post, we'll walk through a practical workflow: download your strings, enrich them using an AI agent with the help of Crowdin Skills, and push the result back to Crowdin.

The Workflow​

A string like Name or Post means something very different depending on where it appears in your app. Is it username or post title? Is it a label or a button? Without context, translators must guess - and guesses lead to mistakes.

The context enrichment workflow with Crowdin CLI is straightforward:

1. Download strings context to a local file
2. Enrich the context using an AI agent
3. Upload the enriched context back to Crowdin

Let's walk through each step.

Step 0: Install Crowdin Skills​

Crowdin Skills is a collection of agent skills that teach your AI agent how to work with Crowdin - including how to correctly fill string context.

Install all Crowdin skills with a single command:

npx skills add crowdin/skills

This installs two skills relevant to context enrichment:

• crowdin-context-cli - documents the context download and context upload CLI commands so the agent understands the full workflow and the JSONL format.
• context-extraction - teaches the agent how to write good context descriptions: which strings need context (ambiguous short words, plurals, UI labels with inline formatting), how to write concise 1–3 sentence descriptions that explain the UI element type and placement, and what not to touch in the file.

Step 1: Download Strings Context​

The context download command exports 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 in the current directory. You can also customize the output path with the --to option. You don't need to execute this command manually. There is no need to execute this command manually. The agent can do it for you when you ask it to.

Filtering Strings​

You rarely need to enrich all strings at once. The command supports a range of filters so you can focus on exactly what matters:

These filters are especially useful when running context enrichment as part of a CI/CD pipeline or when incrementally updating context for new strings.

Step 2: Enrich Context with AI Agents​

Your AI agent already knows your codebase, so it can write the best description possible for your strings. Once you have your crowdin-context.jsonl file, it's time to let the agent do the heavy lifting.

Step 3: Upload Enriched Context​

Once your AI agent has enriched the crowdin-context.jsonl file, upload it back to Crowdin with the context upload command:

crowdin context upload

Use the --from option if you used a custom path with --to during download.

Additional Commands​

Check Context Coverage​

The context status command gives you a quick overview of how much of your project has context filled in:

crowdin context status
✔️  Fetching project info    
Context Status for Project "Demo" (ID: 00000)

Total strings:       3484
With AI context:     50 (1.44%)
Without AI context:  3434 (98.56%)
With manual context: 3484 (100.00%)

Use --by-file to see a breakdown per source file - useful for spotting which files need the most attention.

Reset AI-Generated Context​

The context reset command removes AI-generated context from strings while preserving any manually written context:

crowdin context reset --all

You can scope the reset to specific files, labels, or branches using the same filtering options available on the other commands. The --all flag is required when no filter is specified, as a safety measure against accidental bulk resets.

Summary​

This post showed how to enrich your project's string context using Crowdin CLI and an AI agent. The workflow: install Crowdin Skills, download strings to a local JSONL file, ask your agent to fill in context, then upload the result back to Crowdin. Use context status to check coverage and context reset to remove AI-generated context when needed.

It has been almost five years since the release of Crowdin CLI 3.0. During this time, we have received extensive feedback from our users and worked diligently to improve the tool. As a result, we've made many changes and improvements.

The new version of Crowdin CLI introduces a lot of new features and enhancements. We've added new commands, updated existing ones, and significantly improved the overall user experience.

This post is intended to highlight the most important changes and help you migrate from CLI 3.x with ease. It is not a comprehensive guide, but rather a list of the most important changes.

Requirements​

As part of this release, the minimum Java version has been updated to 17 LTS.

New commands​

• crowdin config
• crowdin project
• crowdin language
• crowdin branch clone
• crowdin branch merge
• crowdin completion

New command options​

Command updates​

Pre-translate​

We've made significant changes to the pre-translate command. The command now requires the --file parameter for file-based projects. The --file parameter determines which files to pre-translate. Previously, the command relied on the current configuration, and it was only possible to pre-translate the files that matched the configuration patterns. Now you can specify any file(s) in the Crowdin project to pre-translate.

Tip: Use the file list command to get the file paths in the current project.

String​

• The string edit command changes:
  • Moved id to a parameter instead of a command option.
  • Now --identifier is used as an option to edit and not to find the needed string.
• The string delete command changes:
  • Moved id to a parameter instead of a command option. Now only 1 string can be removed at a time.
  • Removed the --text and --identifier options. Only ID is used as a string identifier.
• The comment command has been replaced with the comment add command.

-crowdin string add Download --file main/src/strings.xml --identifier "download"
+crowdin string add Download --file src/strings.xml --branch main --identifier "download"

-crowdin string edit --id 12 --text new
+crowdin string edit 12 --text new

-crowdin string delete --id 12
+crowdin string delete 12

List​

• The list branches command has been replaced with the branch list command.
• The list sources command has been replaced with the config sources command.
• The list translations command has been replaced with the config translations command.
• The list project command has been replaced with the file list command.
• The list languages command has been replaced with the language list command.

-crowdin list branches
+crowdin branch list

-crowdin list project
+crowdin file list

-crowdin list sources
+crowdin config sources

-crowdin list translations
+crowdin config translations

-crowdin list languages
+crowdin language list

Glossary​

The --name option has been removed from the upload and download commands, and the <id> parameter is now used instead:

-crowdin glossary download --id <id>
-crowdin glossary download --name <name>
+crowdin glossary download <id>

-crowdin glossary upload <file> --name <name>
+crowdin glossary upload <file> --id <id>

Tip: Use the glossary list command to get the id.

TM​

The --name option has been removed from upload and download commands, and the <id> parameter is now used instead:

-crowdin tm download --id <id>
-crowdin tm download --name <name>
+crowdin tm download <id>

-crowdin tm upload <file> --name <name>
+crowdin tm upload <file> --id <id>

Tip: Use the tm list command to get the id.

Screenshot​

The --name option has been replaced with the <id> parameter for the delete command because there can be multiple screenshots with the same name:

-crowdin screenshot delete <name>
+crowdin screenshot delete <id>

Tip: Use the screenshot list command to get the id.

Init​

The generate alias has been removed:

-crowdin generate
+crowdin init

Lint​

The lint command has been replaced with the config lint command:

- crowdin lint
+ crowdin config lint

Bundle​

The download bundle command has been replaced with the bundle download command.

Exit codes​

This release also includes more exit codes (previously we only had 0 for success and 1 for failure).

See the official documentation for more details.

Environment variables​

Previously, it was necessary to specify the environment variables in the configuration file in order to load the configuration from them.

Now the CLI can automatically pick up the environment variables if they are set in the shell. The supported environment variables are:

• CROWDIN_PROJECT_ID
• CROWDIN_PERSONAL_TOKEN
• CROWDIN_BASE_PATH
• CROWDIN_BASE_URL

Thanks to this change, it's now easier to work with the CLI in CI/CD environments, especially in the no-configuration-file use case.

Visit the Configuration File article for more details.

Deprecations and backward compatibility​

Download targets​

The deprecated download targets command has been removed. The configuration file no longer supports the targets section as well. Use Bundles instead.

Task command​

Removed backward compatibility for file numeric id in the task add command. Now only file path is supported:

-crowdin task add --file 12 --language uk
+crowdin task add --file src/strings.xml --language uk

Command output updates​

In this release, we've also reviewed and improved the output of many commands to make them more informative and user-friendly. Changes include list, add, upload, download, delete, and other commands for various resources.

There are several commands you can use to manage your files. The CLI provides high-level commands that are easy to use and low-level commands that give you more flexibility and control.

The high level commands are crowdin upload and crowdin download. These commands are easy to use and they rely mostly on the configuration file.

The low-level commands include crowdin file upload, crowdin file download, and crowdin file delete. These commands are more like a lightweight wrapper over the Crowdin API. They give you more flexibility and control over the file management process.

The upload and download commands​

Configuration​

These commands are usually being used along with the configuration file.

For example, you have a project with the following structure:

.
├── locales
│   ├── en
│   │   └── messages.xml
│   └── fr
│   │   └── messages.xml
│   └── de
│   │   └── messages.xml
│   └── uk
│       └── messages.xml
└── ...

Let's create a simple configuration file:

crowdin init -s "locales/en/*.xml" -t "locales/%two_letters_code%/%original_file_name%"

This command prompts you for the project identifier and API token. It will then create a configuration file with the following content:

"project_id": "<your-crowdin-project-id>"
"api_token": "<your-personal-access-token>"
"base_path": "."
"base_url": "https://api.crowdin.com"

"preserve_hierarchy": true

files: [
  {
    "source": "locales/en/*.xml",
    "translation": "locales/%two_letters_code%/%original_file_name%"
  }
]

Let's take a closer look at some of the configuration options.

The source and translation configurations represent the patterns for the source and translation files. The source configuration is a glob pattern pointing to the source files. The translation configuration is a path to the translation files (existing ones or where to put downloaded translations). The %two_letters_code% and %original_file_name% are the placeholders that will be replaced by the language code and the original source file name, respectively.

Upload content​

Let's upload the source files and existing translations to Crowdin:

crowdin upload sources
crowdin upload translations -l fr

We can also check the translation and proofreading progress before downloading the translations:

crowdin status

Download content​

When the translations are ready, we can download them:

crowdin download translations

The above command will download all the translations to the path specified in the translation configuration.

To download the source files from Crowdin, run the following command:

crowdin download sources

The file command​

The file command is a low-level command that allows you to manage your files with more flexibility and control.

This command does not require a files section in the configuration file, or a configuration file at all. You can specify all the necessary parameters in the command itself.

Upload content​

Let's use the same project structure as in the previous example and try to upload the source files and existing translations to Crowdin:

crowdin file upload locales/en/messages.xml -d messages.xml

The command above will take the file locales/en/messages.xml and upload it to Crowdin as messages.xml in the root directory. The -d/--dest option stands for destination and allows you to specify the path to the file in Crowdin. If you omit the -d option, the file will be uploaded to the root directory with the same path as the source file.

You can upload strings to string-based projects as well:

crowdin file upload locales/en/messages.xml --branch main --cleanup-mode --update-strings

The --branch option is required for string-based projects. The --cleanup-mode and --update-strings options are optional. Use the --update-strings option to update strings with the same identifiers and the --cleanup-mode option to remove strings that are not present in the uploaded file.

To upload translations you need to specify the -l/--language option:

crowdin file upload locales/fr/messages.xml -d messages.xml -l fr

The -d/--dest option is the path to the file in Crowdin where the translations should be uploaded. -l - Crowdin language code.

Visit the crowdin file upload page to see all the options available.

Download content​

Using the crowdin file download you can download either the source files or translations. Let's try to download the source file from Crowdin:

crowdin file download messages.xml -d locales/en/messages.xml

The above command will download the messages.xml file from Crowdin and store it in the locales/en/messages.xml path.

To download translations, you need to specify the -l/--language option:

crowdin file download messages.xml -d locales/fr -l fr

The -d/--dest option here is the path to the file on your local machine where the translations should be stored. -l - Crowdin language code.

Visit the crowdin file download page to see all the available options.

crowdin file upload locales/en/messages.xml \
  -d messages.xml \
  -i <your-project-id> \
  -T <your-personal-access-token> \
  /

Further Reading​

• Configuration
• Advanced Usage
• CI/CD Integration
• FAQ