Trados

Connect AI agents to RWS Trados Cloud, the platform behind Trados Enterprise and Trados Team. Agents create localization projects, upload source files, run analysis and quotes, assign tasks, and query translation memories and termbases.

74 actions Custom auth available

Platform Documentation ↗

A schedule polls, because Trados webhook subscriptions can only be created in the Trados interface and this connector ships no triggers

List Projects with a Last Modified From filter returns the projects that moved since the previous run

List Target Files reports which languages have reached a finished status

List Target File Versions resolves the version type first, because the same download returns either the native document or the bilingual BCM file

Download Target File Version stores the translated file in FlowRunner file storage and returns a URL the flow can use

The localization owner receives the delivered languages and the post-edited segments waiting to go back into the memory

The owner confirms the target memory and the overwrite behaviour before Import Translation Memory runs

What This Integration Enables

Trados Cloud is not a translation API, and treating it as one is the fastest way to build the wrong flow. It is the project and asset system that sits behind human translators: projects that carry files through a workflow, tasks that real people accept and complete, quotes that get approved before work starts, and the linguistic assets that make the next project cheaper than the last. The connector reflects that shape. Create Project, Add Source File, and Start Project launch work. List Project Tasks and Assign Task move it. Export Quote Report prices it. Translation memories and termbases hold everything that carries forward.

The durable asset is the translation memory, and this is the point most localization automation misses. A project is temporary. A memory is not. Look Up Translation and Concordance Search read against it, Add Translation Unit writes a single approved segment into it, and Import Translation Memory loads a whole TMX file into it. Whatever ends up in that memory becomes the leverage on every future project run against the engine that uses it, appearing in every subsequent analysis as a match and reducing every subsequent quote. Which means it also becomes the wording your company ships, in every language, until somebody notices.

Files move in both directions. Uploads take a file URL rather than raw bytes, so a FlowRunner file storage URL or any publicly reachable URL works, and the connector fetches the content and posts it to Trados as multipart form data. Downloads land in FlowRunner file storage and return a URL. One detail saves real debugging time: Trados sends the original file name in a response header, and FlowRunner exposes response bodies rather than headers, so the original name cannot be recovered. Set the File Name parameter when the name matters. Another: zip archives added through Add Source File are always stored as reference files rather than translatable content, so use Upload Zip File with Attach Source Files to add an archive's contents as translatable.

Exports, imports, and file analysis are all start then poll. The start action returns an identifier, a Poll action reports status, and only then can the result be downloaded. Project file exports and quote report exports expire twenty minutes after the export was started, not twenty minutes after packaging finishes, so a flow that polls lazily gets a 404 rather than a file. File analysis results are discarded after twenty four hours.

There are no triggers. Trados Cloud does send webhooks, but subscriptions can only be created in the Trados interface on the Webhooks tab of a custom application, and there is no webhook subscription endpoint in the public API. Poll instead: List Projects with a Last Modified From filter, List Project Tasks, List My Tasks, or List Target Files watching for a finished status. Rate limits shape how hard a flow can poll, and Create Project and Export Quote Report carry their own tighter limits, so read the live values with List Rate Limits rather than assuming the published defaults.

One further constraint deserves attention before a flow is designed rather than after. The connection acts as a service user and inherits that user's group roles and folder location. Anything the service user cannot see through the account hierarchy is invisible to the flow, and anything it can change, the flow can change. Give the integration its own service user in the narrowest folder that covers the work, and note that swapping the service user on an existing application takes ten to twenty minutes to propagate, during which calls may use either identity.

Without FlowRunner

Projects created by hand Every localization job is set up in the Trados interface from the last one, by memory

Delivery is a checking exercise Somebody logs in on Friday to see which languages finished this week

Post-edited work is lost Corrections a reviewer made in a document never make it back into the memory

With FlowRunner

Projects created from approved content The source files, template, and due date come from the system where the content was approved

Finished languages collect themselves Completed target files are downloaded and pushed back into the CMS or repository they came from

Approved translations become reusable Post-edited segments are written back so the next project leverages them instead of paying again

Use Case Scenarios

Approved content becomes a running project without anyone logging in

A release note is approved in Contentful. The flow creates the Trados project from a project template, which supplies the translation engine, file processing configuration, and workflow the project needs before it can start, then calls Add Source File for each document and Start Project to launch the workflow. Template values take precedence over anything set directly on the project, which is worth knowing when a due date appears not to have applied. If translatable files are added later, Start Project is called again so the new files enter the workflow rather than sitting outside it. The project identifier is written back onto the content record so every later step in this and other flows can find it.

Quote before commitment

The expensive mistake in localization is starting work before anyone knows what it costs. This flow inverts the usual order. It calls Upload Zip File with the candidate content, polls with Poll Upload Zip File, then calls Request File Analysis against a pricing model and polls with Poll File Analysis to get a word count and an estimated cost before a project exists at all. The figure goes to the content owner in Slack with the language list and the deadline. Only on approval does the flow call Create Project. For projects that already exist, Export Quote Report followed by Poll Quote Report Export and Download Quote Report produces the same conversation with a formal document attached, remembering that the export expires twenty minutes from when it started.

Delivery that lands where the content lives

A polling flow calls List Target Files for each active project and watches for a finished status. For each finished language it calls List Target File Versions to determine the version type. That step is not optional: Download Target File Version returns either the native document or the bilingual BCM file depending entirely on which version identifier is passed, and a flow that guesses will eventually push a bilingual working file into a production CMS. It downloads the right one, stores it, and pushes it back into Dropbox or the repository the source came from. When every language on a project is delivered, Complete Project closes it and a summary goes to the requester. Nobody logged in to check.

Human-in-Loop Highlight

Import Translation Memory is the gate, and it is the clearest case of propagation in this entire connector. Trados publishes no endpoint that rolls an import back. The If Target Segments Differ option decides whether existing translations are overwritten by the incoming ones, and that choice is made in the request, before the import, not corrected afterwards. So an import into a production memory does two things at once: it adds whatever the TMX contains, and it can silently replace segments a human translator approved months ago with segments nobody has read. Then the consequence compounds rather than sitting still. Every future project run against a translation engine using that memory will surface the imported segments as matches. They will appear in the analysis, reduce the word count, lower the quote, and be presented to the next translator as previously approved wording, which is the strongest possible signal to leave them alone. A single unreviewed import becomes the company's official phrasing in that language, and the mechanism that makes translation memory valuable is precisely the mechanism that spreads the error. It is also the reason the Resource Strategy setting matters: a project bound to the live shared resource rather than a private copy inherits later edits to that resource while it runs.

So the agent goes right up to the edge and stops. It resolves the target memory through List Translation Memories, reports how many segments the import contains and how many of those collide with existing entries, states which overwrite behaviour is set, and asks: "Importing 4,180 segments of post-edited German into TM Marketing DE-DE. 612 segments already exist with different targets. If Target Segments Differ is set to overwrite. These will replace approved translations and will be offered as matches on every future project. Proceed, switch to keep existing, or cancel?" A localization owner answers, and the import runs on that answer. Everything cheaper and narrower stays automatic: Add Translation Unit writes one approved segment at a time and is a natural fit for a post-editing loop where a human already approved the string. This is human-in-the-loop sized to reach rather than to effort, because the write that takes one call is the write that changes every project after it.

Agent processes routinely

Detects exception requiring judgment

Clear match Continues automatically

Ambiguous Routes to human via preferred channel

Human decides

Agent resumes with decision

Agent Capabilities

74 actions

Account

List Regions Returns the current set of Trados regions and needs no credentials. Useful because RWS asks integrations not to hardcode hosts, since new regions are added over time.
Get My User Returns the service user the connection acts as. Worth calling once, because that user's group roles and folder location define everything the flow can see and change.
List My Accounts Lists the accounts available to the connection.
List Rate Limits Returns the live rate limits for the account. Read these rather than assuming the published defaults, since RWS raises them per account on request.

Projects

List Projects Lists projects with filters including Last Modified From and project name. The primary polling surface in place of triggers.
Get Project Retrieves one project. Its default representation includes the full project plan, analysis statistics, and quote, so use the Fields parameter to trim the response when only a status is needed.
Create Project Creates a project, which starts out empty. Carries its own tighter rate limit than the account default, so a bulk provisioning flow needs pacing.
Update Project Updates a project using JSON merge patch semantics, so only the values supplied change. Rescheduling by changing the due date is allowed only when the workflow has no Customer Quote Approval task, or at least one such approval has already closed.
Start Project Starts the project workflow. Must be called again after adding translatable files to an already started project, otherwise the new files never enter the workflow.
Complete Project Marks a project complete. The natural closing step once every language has been delivered.
Delete Project Deletes a project with its files, tasks, and analysis. Trados publishes no endpoint that restores it.
Get Project Configuration Returns the project's configuration, including the translation engine, file processing configuration, and workflow it is bound to.

Project Files

List Source Files Lists the source files on a project.
Get Source File Retrieves one source file's details.
Add Source File Adds a source file from a URL, which the connector fetches and posts as multipart content. Zip archives added this way are always stored as reference files, never as translatable content.
Attach Source Files Attaches files from an uploaded zip to a project as translatable content. The companion to Upload Zip File, and the correct route for adding an archive's contents.
List Source File Versions Lists the versions of a source file.
Download Source File Version Downloads a specific source file version into FlowRunner file storage.
List Target Files Lists the target files on a project with their status. Watching this for a finished status is the standard delivery poll.
Get Target File Retrieves one target file's details.
List Target File Versions Lists the versions of a target file, including the version type. Read this before downloading, because the type decides what the download returns.
Download Target File Version Downloads a target file version into FlowRunner file storage. Returns either the native document or the bilingual BCM file, decided entirely by which version identifier is passed.
Export Project Files Starts an export of a project's files. The first half of a start then poll pattern.
Poll Project Files Export Reports the status of a project files export. The export expires twenty minutes after it was started, not twenty minutes after packaging finishes.
Download Exported Project Files Downloads the completed project files archive into FlowRunner file storage, inside that twenty minute window.
Upload Zip File Uploads a zip archive from a URL for analysis or for attaching to a project.
Poll Upload Zip File Reports the status of a zip upload.

File Analysis

Request File Analysis Requests a word count and leverage analysis against a pricing model. The action that lets a flow price work before a project exists.
Poll File Analysis Reports the status of a file analysis. Results are discarded twenty four hours after they are generated.

Tasks

List Project Tasks Lists every task on a project, regardless of assignee. The read that shows where work is actually sitting.
List My Tasks Lists tasks assigned to the connection's service user only. Narrower than it looks, which surprises people building dashboards.
Get Task Retrieves one task, including the outcomes defined on its task type. Read the outcomes here rather than guessing, because they differ per task type.
Assign Task Assigns a task. This replaces the whole assignee list rather than adding to it, so anyone not listed loses the task.
Accept Task Accepts a task, moving it in progress. Normally required before a task can be completed.
Complete Task Completes a task with an outcome that must be one of the outcomes defined on the task's task type.
Reject Task Rejects a task.
Release Task Releases a task back to the pool.

Quotes

Export Quote Report Starts a quote report export. Carries its own tighter rate limit than the account default.
Poll Quote Report Export Reports the status of a quote report export. RWS recommends polling at around twenty second intervals, and the export expires twenty minutes after it started.
Download Quote Report Downloads the completed quote report into FlowRunner file storage.

Customers

List Customers Lists customers on the account.
Get Customer Retrieves one customer.
Create Customer Creates a customer, typically mirroring an account record from a CRM.
Update Customer Updates a customer's details.
Delete Customer Deletes a customer. Trados publishes no endpoint that restores it.

Translation Memories

List Translation Memories Lists the translation memories on the account. The read to run before any write, because it is how a flow confirms which memory it is about to change.
Get Translation Memory Retrieves one translation memory with its language directions and configuration.
Create Translation Memory Creates a new translation memory. Creating a fresh memory for unreviewed content is often the right alternative to importing into a production one.
Import Translation Memory Imports a TMX file into a translation memory. Cannot be rolled back, and the If Target Segments Differ option decides whether existing translations are overwritten, which must be chosen before the import rather than corrected after. The write with the longest reach in this connector.
Poll Translation Memory Import Reports the status of a translation memory import.
Export Translation Memory Starts an export of a translation memory.
Poll Translation Memory Export Reports the status of a translation memory export.
Download Exported Translation Memory Downloads the exported memory into FlowRunner file storage. Exports arrive as gzip compressed TMX.

Translation Lookup

Look Up Translation Looks up a segment against a translation memory. Used to check whether a phrase has already been translated before commissioning new work.
Concordance Search Searches a translation memory for how a phrase has been translated previously. The linguist's tool, and a good way for an agent to answer a terminology question without opening a project.
Add Translation Unit Writes a single approved segment into a translation memory. Narrow enough to run automatically in a post-editing loop where a human has already approved the string.

Termbases

List Termbases Lists the termbases on the account.
Get Termbase Retrieves one termbase.
List Termbase Entries Lists the entries in a termbase.
Get Termbase Entry Retrieves one termbase entry.
Create Termbase Entry Creates a termbase entry, which is how approved terminology gets enforced across future projects.
Delete Termbase Entry Deletes a termbase entry. Trados publishes no endpoint that restores it.
Search Termbase Terms Searches a termbase for terms. Used to check whether an approved term already exists before adding a competing one.

Reference Data

List Languages Returns the languages available on the account.
List Folders Lists folders in the account hierarchy. Folder inheritance applies to everything in Trados, so the Locations and Location Strategy parameters on list actions decide how far a query reaches.
Get Root Folder Returns the root folder of the account hierarchy.
List Project Templates Lists project templates. Template values take precedence over anything set directly on a project, which makes this the most consequential reference list here.
List Workflows Lists the workflows available, meaning the sequences of tasks a project's files pass through.
List Translation Engines Lists the translation engines, each binding a set of translation memories and machine translation settings.
List Task Types Lists task types and their defined outcomes. The reference a flow needs before completing a task with an outcome.
List Users Lists users on the account, with an exact email filter.
List Vendors Lists the vendors available for assigning work.
List Pricing Models Lists pricing models, which is what a file analysis is priced against.
List File Processing Configurations Lists file processing configurations, one of the three things a project needs before it can start.

Frequently Asked Questions

What can FlowRunner do with Trados?

FlowRunner agents can run List Regions, Get My User, and List My Accounts in Trados, plus 71 more actions.

Does connecting Trados to FlowRunner require OAuth?

Trados uses a custom authentication method to connect to FlowRunner.

Can Trados trigger a FlowRunner workflow automatically?

Trados doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.

Start building with Trados

$100 in credits. No card required. Connect in minutes.

Start Building Free Talk to us