Using XTM MCP Server to connect your own agents to XTM Cloud
XTM Cloud includes XTM Agent, our AI agent. It can answer product-related questions and support users in many tasks, such as project prioritization and Linguist assignments. For details, see Working with XTM Agent. However, if you prefer, you can also use XTM MCP Server to connect your own agentic solutions to XTM Cloud.
XTM MCP Server creates a connection between your agents and the XTM platform. Our MCP Server offers a “headless” localization experience in which interaction with XTM Cloud is conversational and declarative. Your end-users can use their natural language to communicate with your in-house AI agents, assistants, or chatbots and get things done in XTM Cloud. For example, they can:
search for projects, get project status overview, or assess project criticality.
manage workflows and users.
get information about Linguists.
identify the best Linguists for a specific workflow step and even assign them to the required workflow steps.
For details about its functionality, see the XTM MCP Server capabilities section below.
Prerequisites
To use XTM MCP Server, these criteria must be met:
Public Cloud instances must be running XTM Cloud 26.1, package number 260421_1732 or a more recent version. To check your XTM Cloud version, see Checking your XTM Cloud version and package number.
For Private Cloud instances, contact your account manager and XTM Support to check which package you need to update to.
To request support
Visit our Support Portal.
You must have a license to use the XTM Cloud API.
You must use an MCP host with support for:
a remote server (such as OpenAI Codex, Claude Code, Cursor, Windsurf, etc.).
API key authentication.
custom headers.
Users who will use your in-house agents via XTM MCP Server must have the XTM Cloud Project Manager role.
Limitations
XTM MCP Server currently supports up to 10 requests per second for each authentication token.
Due to authentication constraints, the XTM MCP Server is restricted to a limited set of agents. For details, contact XTM Support.
To request support
Visit our Support Portal.
For capability limitations, see the XTM MCP Server capabilities section below.
Connecting to XTM MCP Server
Follow the steps below to connect to XTM MCP Server:
Generate an API key. To do so, use the Generate token API endpoint (for details, see Generate token in XTM Connect SDK – REST API documentation). In the request body, enter these details:
client: your XTM Cloud Company name.
password: password of the Project Manager user used for integration.
userId: ID of the Project Manager user used for integration. To find the user ID, you can log in to XTM Cloud as this user. Next, select that user's avatar or the user icon (if you have not selected your avatar) in the top right-hand corner of the screen. Finally, select Download my data. The user ID is mentioned in the downloaded Excel file.
integrationKey:
9re8imm23n59scb20adn41b129dasjjj.
Request body:
{ "client": "string", "password": "string", "userId": "integer", "integrationKey": "9re8imm23n59scb20adn41b129dasjjj" }Once you have your API key, call the XTM MCP Server at
mcp.xtm.ai.Verify if your configuration is correct. To do so, in your selected AI tool chat box, enter this prompt: "Get XTM system info". If the connection to XTM MCP Server has been set up correctly, the response you receive will contain the XTM Server version.
In the mcp.json file, in the .cursor folder, add this snippet:
{
"mcpServers": {
"XTM": {
"url": "https://mcp.xtm.ai/mcp",
"headers": {
"Authorization": "XTM-Basic <API Key>",
"X-XTM-System-URL": "<server_URL>",
}
}
}
}Important
Replace the <API Key> with the API key generated as described in step 1.
Replace the <server_URL> with your XTM Cloud production server URL. For the main XTM Cloud instance, https://www.xtm-cloud.com, the X-XTM-System-URL value will be "X-XTM-System-URL": "https://api.xtm-cloud.com/rest-api".
Next, verify the Cursor configuration.
Information received in response to the "Get XTM system info" prompt if the connection has been successful.
In the config.toml file, in the .codex folder, add this snippet:
[mcp_servers.xtm] transport = "streamable_http" url = "https://mcp.xtm.ai/mcp" enabled = true [mcp_servers.xtm.http_headers] Authorization = "XTM-Basic <API Key>" X-XTM-System-URL = "https://api.xtm-cloud.com/rest-api"
Important
Replace the <API Key> with the API key generated as described in step 1.
In this example, the URL for the main XTM Cloud instance, https://www.xtm-cloud.com, has been used. If required, replace it with your own XTM Cloud production server URL in this code snippet: X-XTM-System-URL = "<server_URL>".
XTM MCP Server capabilities
Note
After successful connection, XTM MCP Server will pass information about available tools and limitations to your in-house agentic solutions.
Information in this section is meant for your end-users, who are interacting with your in-house agentic solutions, so that they have a better understanding of actions they can perform within XTM Cloud.
Users with the XTM Cloud Project Manager role can use XTM MCP Server to perform these tasks:
Project-related actions:
Search for projects by name, status, dates, creator, Project Manager, or customer for which they have been created.
Check full project details, which include project languages, word counts, due dates, and customer information.
Check translation progress for specific projects, such as their job statuses, workflow steps, and due dates.
View project workflow configuration: steps, workflow transition types (also known as blocking types), and due date ranges.
User management:
List all internal users in your XTM Cloud instance.
List all subcontractors (LSPs) in your XTM Cloud instance.
List all users with the Linguist role in your XTM Cloud instance.
Check a Linguist's language pairs and access rights to a workflow step, subject matter, and a customer.
Find the best Linguist for a Job, based on their current workload.
Assignment management:
View current assignments in a project.
Assign Linguists to a project.
Assign subcontractors (LSPs) to a project.
Note
Due to the current API limitations, you cannot check LSP language combinations using XTM MCP Server. You also cannot use it to suggest an LSP for a particular workflow step. Before you use XTM MCP Server to assign an LSP to your project, we recommend that you check their access rights in XTM Cloud. For details, see Adding LSP access rights.
Unassign Linguists from a project.
Unassign subcontractors (LSPs) from a project.
Override current Linguist assignment (replace a currently assigned Linguist with a different one without unassigning them first).
Override current LSP assignment (replace a currently assigned LSP with a different one without unassigning them first).
Project costs:
Get a list of all costs generated for the project so far. This list contains information about the cost IDs, what they are based on (metrics vs. time), and when they were generated. The list can be sorted in ascending/descending order.
Check the cost details, based on cost ID. In response, you get the cost total, currency, per-user cost, and per-target-language cost.
Project metrics:
Get project-wide aggregated metrics. In response, you get the summed core metrics (such as number of characters, words, and segments per match type) and aggregated information per workflow step about progress across all target languages.
Get metrics for a selected project target language. In response, you get the summed core metrics for that language (such as number of characters, words, and segments per match type) and information about progress for each workflow step in that target language.
Get metrics for a selected project job. In response, you get the summed core metrics for that job (such as number of characters, words, and segments per match type) and information about progress for each workflow step in that job.
Connectivity verification:
Check XTM Cloud version.
Check server information.
Important information:
Unless you specify a project name, ID, or date range, the search is limited to projects created in the last 14 days.
By default, only active and auto-archived projects are returned. To include archived or inactive projects, ask for them explicitly.
You don't need to specify an exact project name: partial matching is supported.
If you have a project's ID, its details can be fetched regardless of its project status.
When you assign a Linguist, the server validates their language pair compatibility and required access rights automatically.
Limitations:
It is not possible to assign Linguists or LSPs to projects that are still being analyzed or have been archived, auto-archived, or deleted.
It is not yet possible to suggest or assign user groups.
It is not yet possible to assign Linguists to bundles in a project. They can only be assigned to jobs and workflow steps.