Smartling MCP server for AI agents

Authorize glossary entries matching a filter for translation into one or more locales via existing project workflows. This is an asynchronous operation: the response confirms the request was accepted (with an operation_uid), not that it has completed.

Create a new string issue (question, clarification, quality flag) on a Smartling string. # Use cases - Translator flags an ambiguous source string with a CLARIFICATION sub-type - QA raises a POOR_TRANSLATION issue against a translation - PM creates a CUSTOM source issue with a HIGH severity assignment # Examples - Minimal: project_id, issue_text, issue_type=TRANSLATION, issue_sub_type=CLARIFICATION, string={hashcode, locale_id} - With assignee + severity: also pass assignee_user_uid and issue_severity_level=HIGH

Unassign the user currently assigned to an issue. # Use cases - Remove the current assignee when the issue should not be owned by anyone # Examples - project_id="p1", issue_uid="i1"

Download translated content for a file from Smartling - as a single file for one locale, a zip of all locales for one file, or a zip of multiple files across specified locales. # Modes - SINGLE_LOCALE: one file, one locale - returns the translated file itself - ALL_LOCALES_ZIP: one file, every locale that has translations - returns a zip archive - MULTIPLE_FILES_ZIP: multiple files, each with its own locale list - returns a zip archive (up to 500 file/locale combinations total) # Important Notes - For MADCAP, DITA_ZIP, CSV, XLSX_TEMPLATE, and XCSTRINGS files only, locale_id in SINGLE_LOCALE mode also accepts a comma-separated locale list or "all" to get one combined (non-zip) file with all of those locales. - MULTIPLE_FILES_ZIP returns a "no files found" text response instead of a file when file_filter=PUBLISHED_FILES_ONLY excludes every requested file. - Returns a presigned download URL, not the file content itself. The URL is valid for 15 minutes, so fetch it promptly - re-run this tool to get a fresh URL if it expires.

Edit the body text of an existing issue. # Use cases - Fix a typo in an issue's text - Clarify a previously-vague issue description # Examples - project_id="abc", issue_uid="i1", issue_text="Updated description"

Run an automated Linguistic Quality Assurance (LQA) evaluation on a single translated string and get back any detected quality errors. # What is an LQA evaluation? LQA evaluation checks a translation against its source string using Smartling's linguistic quality rules (e.g. blocklist compliance, omissions, mistranslations) and returns the list of errors found, each with a category and severity. It does not modify the string or record the errors anywhere; it only reports what an LQA scan would flag. # Use cases - Automatically screen a translation for quality issues right after it's produced (e.g. from MT or a translator) - Gate a workflow step transition on the string passing LQA with no critical/major errors - Spot-check a translation on demand without waiting for a scheduled LQA pass # Examples - Minimal: account_id, project_uid, hashcode, source_locale_id="en", target_locale_id="fr-FR", original_string, translation - With job/workflow context: also pass job_uid, workflow_uid, workflow_step_uid so the evaluation is attributed to that job step - With custom metadata: metadata={"integrationType": "TEST"}

Search issues within a single Smartling project, with pagination, sorting, and rich filters. # Use cases - List all OPEN issues in a project, newest first - Find HIGH-severity TRANSLATION issues assigned to a specific user - Count + page through a project's issue backlog # Examples - All open issues: project_id="p1", issue_state_codes=["OPENED"] - By assignee: assignee_user_uid="user-1" - Sorted by creation date: sort_by={items:[{direction:"DESC", field_name:"createdDate"}]}, limit=30

Check whether email addresses already belong to a user, account user, or pending invite in a Smartling account - useful before inviting new users.

Get a file's translation status for every locale, and its last-modified date per locale. Returns two independent results - `status` (per-locale progress and word counts) and `lastModified` (per-locale last-modified dates) - from two separate Smartling API calls. They are not merged together.

Retrieve detailed information about a specific string issue. # What is a string issue? A string issue is a question, clarification request, or quality flag raised against a specific string (source or translation) inside a Smartling project. # Use cases - Inspect the full state of one issue by its UID - Read severity, type, comments count, and assignee on a single issue # Examples - Get an issue: project_id="abc123", issue_uid="issue-uid-1"

Get the audit history of a specific Smartling job. # What is job history? Job history is the chronological audit trail of everything that has happened to a job: creation, status changes, content additions/removals, due date changes, priority changes, and rush request create/remove events. Each entry records the action, when it started/ended, who (if anyone) performed it, and event-specific details. # Use cases - User wants to know who changed a job's due date or priority, and when - User needs to see when content was added to or removed from a job - User wants an audit trail of status changes for a job - User is investigating unexpected job behavior and needs the sequence of events # Examples - Get recent history: Provide project_id and job_uid (returns up to 100 most recent entries) - Page through longer histories: Use offset with limit to retrieve subsequent pages, guided by totalCount in the response # Pagination - limit: Page size, 1-100 (default 100 if omitted) - offset: Zero-based offset into the history (default 0 if omitted) - totalCount in the response tells you how many entries exist in total

Get the full category/severity/error-type breakdown for one LQA schema - resolves human-readable names to the UIDs smartling_record_lqa_error and smartling_update_lqa_error need. Check severityTypeCode first: - NAMED: match severityNames[].name to your desired severity, then pass its severityNameUid to smartling_record_lqa_error/smartling_update_lqa_error - POINT_RANGE: pass a numeric severity_score within severityRange.min-max instead - severityNames will be empty

Get a report tracking how translations were modified over time, including incoming/outgoing translation text, edit distances, translator info, workflow details, and timestamps. Apply as many filters as possible (project_uids, job_uids, user_uid, target_locale_ids, a narrow start_date/end_date) - unfiltered calls can return a very large report.

Get word count report from Smartling in CSV format for efficient analysis. # What is word count report? A word count report provides detailed information about word counts and weighted word counts for translation work. It includes data about projects, jobs, translation resources, agencies, locales, workflow steps, and fuzzy matching profiles. # API Constraints **CRITICAL**: The date range between start_date and end_date must be **less than 1 year (365 days)**. The API will return a validation error if the range exceeds this limit. # IMPORTANT: Apply Filters to Minimize Response Size **Always encourage users to apply as many filters as possible** to keep response size manageable: - **project_ids**: Filter by specific projects (most important filter) - **target_locale_ids**: Filter by specific target locales - **job_uids**: Filter by specific jobs - **user_uids**: Filter by specific translation resources - **workflow_step_types**: Filter by specific workflow steps (use smartling_get_workflow_step_types tool first) - **start_date/end_date**: Use narrow date ranges - **fields**: Request only the columns you need (see field list below) # Recommended Field Selection **Minimize the number of fields in the 'fields' parameter**. Only include what's needed for the specific query: - Minimal: "projectId,projectName,targetLocaleId,wordCount" - With job info: "projectId,projectName,targetLocaleId,jobUid,jobName,wordCount" - With translator: "projectId,targetLocaleId,translationResourceName,workflowStepType,wordCount" # Examples - Find projects with most words in French: {"start_date": "2024-01-01", "end_date": "2024-12-31", "target_locale_ids": ["fr-FR"], "fields": "projectId,projectName,wordCount"} - Find strings in AI Review step: {"start_date": "2024-01-01", "end_date": "2024-12-31", "workflow_step_types": ["AI Review"], "fields": "projectId,jobName,workflowStepType,wordCount"} - Translator productivity: {"start_date": "2024-01-01", "end_date": "2024-01-31", "user_uids": ["user-123"], "fields": "translationResourceName,targetLocaleId,wordCount"}

List the connectors (integrations, e.g. Intercom, Zendesk) available on a Smartling account, to get each connector's normalized integration_id. See the integration-translation-request skill (smartling_load_skill) for the full workflow.

Search a connector's content (e.g. Intercom articles) for assets to translate. See the integration-translation-request skill (smartling_load_skill) for the full workflow, including how to learn valid asset_type/submission_status/custom_filters values first. smartling_locale_id must be one of the Smartling locale codes returned by smartling_integration_get_search_context's localeMappings for this connector - passing a locale id with no mapping fails with an error rather than searching unfiltered.

List glossaries by set of filters. # Use cases - User requests to retrieve glossaries: - User asked to retrieve a list of all glossaries - User asked to search for glossaries by name or description - User asked to filter glossaries by state (active, archived, both) - User asked to look for glossary for a specific target locale - User asked to retrieve several glossaries by their UIDs

List job batches for a Smartling project, optionally filtered by translation job or status. Page through results using offset/limit, guided by totalCount in the response.

Read back the LQA findings already recorded on one specific string.

List the attachments (links and files) on a Smartling style guide.

Load the full instructions for one skill by id (from smartling_list_skills), then follow those instructions directly to complete the task - the returned text is not meant to be summarized or shown to the user verbatim, it is a set of directions for you to act on.

Archive or restore one or more glossaries in bulk.

Create a new glossary entry or update an existing one.

Create, update, or delete a glossary label.

Remove all tags from strings in a Smartling project. # Use cases - Clean slate for string reorganization - remove all tags before applying new categorization - Prepare strings for migration to a different tagging system - Batch clean up of strings during project restructuring # Examples - Clean up strings before project migration - Remove all tags from deprecated feature strings - Reset tagging for strings that will be recategorized - Clean up test strings that accumulated development tags # Parameters - project_id: The unique identifier of the Smartling project - string_hashcodes: Array of string hashcodes to remove all tags from (max 1000 hashcodes) # Important Note This operation removes ALL tags from the specified strings. Use with caution as this action cannot be easily undone.

Requests a short-lived URL for uploading a local file, for use with other tools that accept an "upload_id" parameter (e.g. smartling_machine_translate_file). # Use cases - Preferred, default first step before translating or otherwise processing any local file, whenever you can run shell commands (e.g. Claude Code, Codex CLI, opencode, Cursor, Cline, Windsurf) - use this instead of inline base64 content regardless of file size or type. Only usable by clients that can run shell commands (e.g. Claude Code, Codex CLI, opencode, Cursor, Cline, Windsurf) to PUT the file to the returned URL - not usable from claude.ai or other clients without local file/shell access, which should keep using inline base64 content instead. The returned URL expires in 15 minutes. Upload immediately after requesting it, then pass the returned upload_id to the tool that needs the file.

Replace the target locales bound to a Smartling style guide. This is a full-replace operation, not additive: the style guide's target locales become exactly the list you pass here. Locales not included are unbound; to add a locale while keeping existing ones, include all of them (existing plus new) in locale_ids. Pass an empty array to unbind all target locales.

Mark an issue as answered or un-answered. # Use cases - Mark an issue answered after a reply is posted - Clear the answered flag if the reply is later found inadequate # Examples - project_id="p1", issue_uid="i1", answered=true

Change the type and sub-type of an issue (optionally narrowing to one locale). # Use cases - Re-categorize a CLARIFICATION issue as MISSPELLING after triage - Move a TRANSLATION issue to a specific locale # Examples - project_id="p1", issue_uid="i1", issue_type="TRANSLATION", issue_sub_type="MISSPELLING" - with locale: also locale_id="fr-FR"