The Transifex MCP Server connects your AI tools, such as Cursor, Claude Code, Codex, or Windsurf, directly to your Transifex account. MCP (Model Context Protocol) lets AI tools use tools and information from other systems. An MCP server describes the tools it offers and handles their requests.

Instead of opening a browser, clicking through menus, and copying text back and forth, you can talk to your AI app in plain English. For example, you can tell it to "Show me which projects need translating" or "Create a new AI translation task," and the AI will handle it automatically.

You also have access to the MCP tools via XTM Agent, the AI assistant in the Transifex dashboard.

Transifex dashboard with an arrow pointing to the XTM Agent button, which sits below the Customer support button in the lower-right corner.

What can your AI agent do?

Once connected, your AI agent can interact with Transifex to handle these tasks:

Find Your Work: See all the organizations and projects you have access to.
Check Status: Review project details, identify which languages are being targeted, and see how close the translation is to completion.
Prioritize Tasks: Compare different projects in your organization to see which ones have the most untranslated words, so you know what to work on first.
Manage Tasks: View, create, or modify translation tasks (including setting up automated AI translations).
Assign Self to a Task: Have the AI automatically assign or unassign your account to specific human-translation tasks.
Create a Project: Set up a new project with its source language and target languages.
Upload Files: Send source files and translation files from your computer to a file-based project. A resource that doesn't exist yet is created for you.
Download Files: Save source files and translation files from a file-based project to your computer, in any of the download modes Transifex offers.

Setup requirements

Before you begin, make sure you have:

A Transifex account with access to an organization.
A way to authenticate: either a Transifex API token or your regular Transifex login (see the two options below).
An AI tool that supports MCP connections (like Cursor or Codex).
To upload or download files, an AI tool that can run commands in a terminal on your computer. The server never sees your files' content; your AI tool runs a command that moves each file.

Connecting to Transifex

There are two ways to connect your AI tool to Transifex: API Token or OAuth. Pick whichever is easier for your setup.

Option 1: API token

Generate a Transifex API token by logging into Transifex and following the steps in Generating an API token. Your AI tool sends this token in the following format:

Authorization: Bearer <Your-Transifex-API-Token>

Connection URL:

https://mcp.transifex.ai/mcp

Codex configuration

Open or create a file named config.toml in the .codex folder in your home directory (~/.codex/config.toml), and add the following lines. Replace <Your-Transifex-API-Token> with your actual token. You can also scope the server to one project with a .codex/config.toml file in the project folder. Codex reads project-scoped files only for trusted projects. For more detail, see the Codex MCP documentation.

[mcp_servers.transifex]
url = "https://mcp.transifex.ai/mcp"
enabled = true

[mcp_servers.transifex.http_headers]
Authorization = "Bearer <Your-Transifex-API-Token>"

Cursor configuration

Open or create a file named mcp.json inside your project's .cursor folder, and add this code. Replace <Your-Transifex-API-Token> with your actual token:

{
  "mcpServers": {
    "Transifex": {
      "url": "https://mcp.transifex.ai/mcp",
      "headers": {
        "Authorization": "Bearer <Your-Transifex-API-Token>"
      }
    }
  }
}

Option 2: OAuth

With OAuth, you don't need to generate or manage an API token at all. When your AI tool connects to Transifex for the first time, it will open a browser window and walk you through signing in with your regular Transifex credentials. Once you approve the connection, you're set.

Transifex authorization screen that asks you to allow an application to access your account. It lists the permissions the application needs, with Cancel and Allow buttons.

Codex configuration

Open or create a file named config.toml in the .codex folder in your home directory (~/.codex/config.toml), and add the following lines. You can also scope the server to one project with a .codex/config.toml file in the project folder. Codex reads project-scoped files only for trusted projects. For more detail, see the Codex MCP documentation.

[mcp_servers.transifex]
url = "https://mcp.transifex.ai/mcp"
enabled = true

Then run codex mcp login transifex in your terminal to start the sign-in. You can also select Authenticate in the MCP server list of the Codex IDE extension.

Cursor configuration

Open or create a file named mcp.json inside your project's .cursor folder, and add this code:

{
  "mcpServers": {
    "Transifex": {
      "url": "https://mcp.transifex.ai/mcp"
    }
  }
}

Testing the connection

To make sure everything is working, open your Cursor AI sidebar chat and type:

"List the Transifex organizations I can access"

If everything is set up correctly, the AI will talk to Transifex and reply with a list of the organizations you belong to.

To check the connection in Codex, type /mcp in the Codex terminal UI (TUI) to see your active MCP servers, and check that transifex is listed. Then ask Codex the same question as above.

Speed limits (rate limiting)

To keep the system running smoothly for everyone, the server limits how fast requests can be made using your token:

Normal speed: Up to 25 requests per second.
Short bursts: Up to 50 requests at once for brief moments.

Reference library: what tools does the AI have?

When you ask your AI assistant to do something, it chooses from a specific list of built-in tools. Here is a plain-language breakdown of what the AI can call on your behalf.

Organizations and projects

Tool	What it does
`get_organizations`	Lists all your Transifex organizations. The AI uses this if it doesn't know where to look yet.
`get_organization_overview`	Looks at an entire organization to compare projects and sort them by how many words are left to translate. This is useful for spotting what needs urgent attention.
`get_projects`	Shows a quick summary list of projects (names, default languages, and descriptions).
`get_project`	Shows the complete settings for a specific project, including settings such as Translation Memory or AI automation.
`get_project_overview`	Gives a full summary of a single project, showing target languages, recent file updates, and translation percentages.
`create_project`	Creates a new private project, either file-based (strings come from files) or Native (strings come from code through a Transifex Native SDK). You provide the organization, a name, the source language and at least one target language. All other settings keep their defaults.

📝 Note: create_project never creates anything on the first call. The AI shows you a summary and asks you to confirm. Only after you say yes does it create the project. The confirmation expires after 30 minutes.

Files, languages and progress

Tool	What it does
`get_resources`	Lists the files (resources) that exist in a specific project.
`get_resource`	Shows detailed information about a single file, such as its total word and sentence counts.
`get_languages`	Shows all the languages this project is being translated into.
`get_strings`	Looks up the system ID (hash) for each text block. It returns system IDs, not the actual text.
`get_project_language_stats` and `get_resource_language_stats`	Show progress counts (how many words are translated, reviewed, proofread, or untranslated) for an entire project, or for a single file.

Uploading and downloading files

These tools work with file-based projects. Native projects aren't supported.

Tool	What it does
`upload_source`	Sends source files from your computer to resources in a project. If a resource doesn't exist yet, it's created. For each file you can choose whether to replace source strings that were edited in Transifex with the text from the file, and whether to keep the translations of strings whose source text changed.
`upload_translation`	Sends translation files from your computer to existing resources, one file per resource and language. The language must be one of the project's target languages. Transifex decides what the file changes.
`download_file`	Saves a source file or a translation file from a project to your computer. For a translation file, you can choose a download mode (see below).
`get_transfer_status`	Checks on earlier uploads and downloads. It shows whether a file is waiting for its upload, still processing, done or failed. For a finished upload, it shows how many strings were created, updated, deleted or skipped. For a failed one, it lists the lines in the file that Transifex couldn't parse.

How file transfers work

Files travel directly between your computer and Transifex. Your AI tool never reads the content of the file. Instead, the server gives it a one-time command for each file, and your AI tool runs that command in your terminal.

Uploads need your approval. upload_source and upload_translation change nothing on the first call. The AI shows you a summary of each file, and sends the files only after you say yes. The confirmation expires after 30 minutes.
Each command works once. The link in a command expires after 10 minutes. If it expires, ask the AI to try again to get a new one.
You can check the result. A transfer ID stays valid for 24 hours. Ask the AI to check the transfer status after the command has run.
Downloads only read. Downloading a file changes nothing in Transifex. If Transifex is still preparing a file, the AI waits and runs the command once the file is ready.
Limits. You can transfer up to 25 files in one request.

⚠️ Warning: Uploading a source file can change your project. With the option to keep translations switched off, translations of strings whose source text changed are removed. With the option to replace edited strings switched on, source edits you made in Transifex are overwritten by the file's text. Check the summary before you confirm.

Download modes

When you download a translation file, you choose a mode. It decides which translations the file holds. If you don't choose a mode, the AI uses default.

Mode	What the file holds
`default`	Every translation. This is the file you use in your app.
`reviewed`	Only reviewed translations.
`proofread`	Only proofread translations.
`translator`	A file for translating offline, with untranslated strings left empty.
`untranslated`	Only the strings that have no translation yet.
`onlytranslated`	Only the translated strings.
`onlyreviewed`	Only the reviewed strings.
`onlyproofread`	Only the proofread strings.
`sourceastranslation`	Every translation, with the source text in each untranslated string.

📝 Note: The file format decides what happens to strings that a mode leaves out. Depending on the format, they become source text or empty values, or they're left out of the file. Check the file format documentation to know what export to expect.

Managing work and assignments

Tool	What it does
`get_tasks` and `get_task`	List all tasks, or inspect one task (including automated AI jobs), showing due dates, priorities, and who is currently assigned to them.
`get_task_subtasks`	Breaks down a task to show the individual steps required for different languages.
`create_task`	Builds a new translation task. It can set up human translation, human review, or fully automated AI translation options.
`edit_task`	Updates a task's name, description, priority rank, or due date.
`set_my_task_assignment`	Adds you to or removes you from a specific task step.

📝 Note: The server derives who you are strictly from your login (whether that's your API token or your OAuth session). The AI can assign or unassign your account to steps, but it cannot use these tools to assign work to other users.

💡 Tip

Looking for more help? Get support from our Transifex Community Forum!

Find answers or post to get help from Transifex Support and our Community.

What can your AI agent do?

Setup requirements

Connecting to Transifex

Option 1: API token

Codex configuration

Cursor configuration

Option 2: OAuth

Codex configuration

Cursor configuration

Testing the connection

Speed limits (rate limiting)

Reference library: what tools does the AI have?

Organizations and projects

Files, languages and progress

Uploading and downloading files

How file transfers work

Download modes

Managing work and assignments

💡 Tip

Related Articles