W44 – Set up engines, bulk-edit glossaries, and read quality measurements in MCP

The Lingo.dev MCP server can now provision a localization engine from your website, update or delete up to 100 glossary terms per call, read translation quality measurements, and work across every organization you belong to from one connection.

Provision a localization engine through MCP#

Give your AI assistant your website URL, and it provisions a localization engine with automatic context provisioning. You can:

Create and fill a new engine. Lingo.dev reads your site, generates a glossary, a ruleset, and a brand voice, and attaches them to a new localization engine.
Fill an existing engine. Generate only the types of context the engine is missing, or add a new glossary or ruleset next to the current ones. Existing context isn't edited or replaced.
Use text as a source. Send text such as product copy, a style guide, or a term list along with links. A job accepts up to 10 sources.
Follow progress. Provisioning runs in the background and takes a few minutes. Check the job, see each type of context as done, skipped, or failed, and retry the parts that failed.

Update or delete up to 100 glossary terms per MCP call#

Each glossary term update or deletion used to take its own MCP call, which made large glossaries slow to maintain. Now you can:

Update terms in batches. Change the source text, description, or translations of up to 100 terms at once.
Delete terms in batches. Remove up to 100 terms, with all their translations, at once. Deleted terms can't be restored.
See what didn't apply. Each term is processed on its own. If some fail, for example because of an unknown term ID or a conflict with an existing term, the successful updates and deletions are applied, and the failed ones are reported.

For more than 100 terms, ask your AI assistant to use batches of up to 100.

Read translation quality measurements through MCP#

Use the MCP server to inspect existing translation quality measurements. You can:

See what each measurement covers. List measurements with their engines and languages.
Find low-scoring translations. Filter results by engine, language, or measurement.
Understand a score. Read the explanation behind any result.
Track quality over time*. Get pass rates and average scores by language, engine, or model.

Use one MCP connection across all your organizations#

One MCP connection now covers every organization your Lingo.dev account belongs to. You no longer need to disconnect and reconnect the MCP server to work in a different organization.

Choose the organization in the conversation. Your AI assistant lists your organizations and asks which one to use. If you belong to only one, it uses that one.
Keep your permissions. The MCP server acts as you, with your role and permissions in each organization.
Sign in with SSO where it's required. Organizations that require single sign-on are available after you sign in to Lingo.dev through your company's identity provider and reconnect.

Connect the Lingo.dev MCP server to get started.

* Available on the Enterprise plan.

The legacy CLI doesn’t support all the features of the new CLI, including your engine’s configured localization pipeline. Reach out to the Lingo.dev team for help migrating.