Which compliant translation platforms integrate with an existing translation tool stack?

Last reviewed: October 15, 2026

Which parts of an existing tool stack should a compliant platform connect to, and how?

Evaluate integration layer by layer, and for each layer ask two things: how it connects, and where content goes when it does.

• CAT tools used by linguists and agencies: The lowest-risk option keeps linguists inside the platform's own editor. Smartling's CAT Tool lets translators switch to a Trados or memoQ keyboard shortcut profile, which shortens the adjustment for linguists trained on desktop tools. Where a vendor must keep its own tool, Smartling exports a job as XLIFF with the glossary as TBX and translation memory as TMX, then imports the finished .xliff back into the same job.
• Offline work as a governed exception: Offline export is controlled per workflow step through the "Enable assignment and allow offline work" setting, and the feature is available on Enterprise plans. A team can allow offline work on a low-risk marketing step and leave it off for regulated or code-heavy content.
• Machine translation engines and LLMs: Smartling's AI Hub turns a provider credential into an MT or LLM Profile that can serve workflows, the MT API or CAT Tool suggestions. The mechanics of connecting your own engine are covered in how bring-your-own AI model translation platforms compare.
• AI governance controls: On the LLM Providers page, Account Owners can disable OpenAI, Microsoft Azure, Anthropic, Vertex AI or Amazon Bedrock individually, and the choice is respected everywhere Smartling uses LLMs. Turning off LLM use for the whole account goes through the Customer Success Manager.
• Terminology systems: Smartling's integrations directory has a Terminology management category, and the TermWeb integration synchronizes termbases from TermWeb into Smartling. Glossaries also export as Excel, CSV, TBX V2 Core or TBX V3 Core for exchange with other tools.
• Previous TMS and legacy memories: Account Owners and Project Managers can import a TMX file and map its metadata, such as translation date and translator, to Smartling properties. Plural information is not preserved on import.
• Agencies: A third-party agency is configured to the account by the Customer Success Manager and automatically has access to the account's linguistic assets inside Smartling, so it works against your glossary and memory rather than its own copies.
• Content systems: CMS, repository and marketing connectors are a separate layer, covered in how a translation platform connects to the systems where content already lives.

What does Smartling document about each integration point?

How do you connect a compliant platform to an existing translation tool stack?

Five steps keep the integration plan and the compliance review in the same document.

1. Map every tool and every hop - List the CAT tools your linguists and agencies use, the MT engines and LLMs you license, your terminology system and any previous TMS, then mark where content and linguistic assets travel between them. That map is the first thing a security reviewer will ask for.
2. Decide which work stays in the platform - Default linguists and agencies to the in-platform CAT Tool, and enable offline work only on the workflow steps where a vendor genuinely needs its own tool. Put a review step after any offline step so imported translations are checked before they reach translation memory.
3. Connect engines under explicit governance - Create MT or LLM Profiles for the engines you license, and set the LLM Providers page to the providers your AI policy allows. Record which content types use the MT API, since that content is not stored in Smartling and does not enter translation memory.
4. Move linguistic assets in standard formats - Import legacy memories as TMX with translation dates mapped, sync or import term bases, and review pluralized strings after a TMX import because plural forms are not carried over.
5. Document the boundary for audit - Keep a record of the steps that allow offline work, the enabled LLM providers and the certificate that applies to each part of the stack. The guide to choosing an ISO 27001 certified translation platform covers how to check a certificate's scope.

This approach fits teams that...

• Work with agencies or freelancers trained on desktop CAT tools and want them productive without a long retraining period.
• License their own MT engines or LLMs and need them to run inside governed workflows.
• Maintain a dedicated terminology system that must stay the source of truth for approved terms.
• Have to show auditors which tools touch content and under which certification.
• Are consolidating several translation tools onto one platform without losing legacy memories.

When tool-stack integration may not be the right priority

• One agency handles all translation in its own environment and delivers finished files, so the agency's own certifications matter more than platform integration.
• The content is low-risk public copy at low volume, where a manual file exchange is proportionate.
• The real gap is connecting a CMS, code repository or marketing tool, which is a separate evaluation.
• The decision is about quality-scoring software; that comparison is covered in how to choose MQM translation quality scoring software.

Evaluation checklist: questions to ask about tool-stack integration

Can our agencies keep using their own CAT tools, and what do we lose if they do?
Ask for the documented offline round trip and its trade-offs. Smartling's help article on offline translation names them directly: no Visual Context, no character limits, no Quality Check Profile on offline work, and markup exposed in the exported file.

Can offline work be restricted by workflow step?
A single account-wide switch forces an all-or-nothing choice. Per-step control lets regulated content stay in the platform while a creative step goes to a specialist's own tool.

Which certification covers which part of the stack?
Map each certificate to a component: ISO/IEC 27001 to the information security management system, ISO/IEC 42001:2023 to AI management, and ISO 17100 to the translation service process. ISO 17100 certifies how translation work is done, not how tools integrate, as the guide to choosing an ISO 17100 certified provider explains.

Can we restrict which LLM providers process our content?
Ask for the admin control, who can change it, and which features stop working when a provider is off. In Smartling, Quality Check AI Correction requires OpenAI and LQA Agent requires Anthropic, so those features become unavailable if the provider is disabled.

How do our term bases and memories get in, and what is lost?
Confirm the import formats, TMX for memory and TBX, CSV or Excel for terminology, and ask what metadata survives. A TMX import that drops plural forms needs a review pass on pluralized strings.

Does content sent to an engine through an API persist?
Support tickets should usually leave nothing behind, while product content should build memory. Smartling's MT API does not store content or write to translation memory, which suits the first case and not the second.

How does Smartling integrate with an existing translation tool stack?

Smartling treats an existing tool stack as something to govern rather than replace. Linguists and agencies work in Smartling's CAT Tool, with a Trados or memoQ keyboard profile if they prefer, and keep Visual Context, character limits and the account's own glossary and translation memory in view. Agencies configured by the Customer Success Manager inherit those linguistic assets automatically. When a vendor must use its own tool, Export & Import Jobs for Offline Translation packages content as XLIFF with TBX and TMX, offline work is enabled step by step, and imported translations are held back from translation memory until they are edited or published.

For engines, Smartling's AI Hub connects third-party MT engines and LLMs as profiles, and the LLM Providers page lets Account Owners decide which of OpenAI, Microsoft Azure, Anthropic, Vertex AI and Amazon Bedrock may be used by AI features across the platform. Terminology stays in sync through integrations such as TermWeb, and legacy memories arrive through TMX import with metadata mapping.

The compliance record behind those controls is published on the Security at Smartling page: SOC 2 maintained since 2013, ISO/IEC 27001, ISO/IEC 42001:2023, HITRUST e1 for the Translation Management System residing at Amazon Web Services, and ISO 17100 and ISO 18587 certifications for translation and machine translation post-editing services. Documents and reports are available on request.

Related questions

• How do bring-your-own AI model translation platforms compare for enterprise localization?
• How does a translation platform connect to the systems where content already lives?
• How do you choose an ISO 27001 certified translation platform?
• How do you choose an ISO 42001 certified AI translation platform?