Draft — not yet reviewed
A lawyer has not reviewed this text yet, and it may change. Placeholders in [brackets] will be filled in before launch.
Legal
Where your data goes
Updated 2026-11-08
Which text leaves Figma, when, and which services process it for Mondoword: AI providers, spell check, hosting, database, sign-in and integrations.
Mondoword reads text from your Figma file so it can store and translate it. This page lists what leaves Figma, where it goes, and how long it stays. It is part of our Privacy policy.
What leaves Figma
| When you… | What is sent | Where it goes |
|---|---|---|
| Open a project in the plugin | The source-language text of the project's keys, for spell check | LanguageTool, straight from the plugin |
| Create, link or edit keys | Key names, texts, and Figma file and layer IDs | Our servers and database |
| Translate, or use Auto-link | The text to translate or link, key names, target languages, glossary terms, style guide, and the layer names and text of the frame it sits in | Our servers, then an AI provider |
| Link a layer in a new frame | The frame's layer names and text, to write a short summary of the screen | Our servers, then Google Gemini |
| Connect Crowdin | Keys, source texts and translations, in both directions | Crowdin |
| Connect an AI agent | Whatever you allow in Agent access | The agent you connected |
We never send images of your designs, coordinates, fonts, or layers you have not linked or selected.
Services that process data for us
| Service | Company | What it receives | Where | Why |
|---|---|---|---|---|
| Claude API | Anthropic | Text for translation and Auto-link (see above) | United States | Default AI for translation and Auto-link |
| Gemini API | Same as Claude when Gemini is chosen or Claude fails; frame text for summaries; project texts for glossary and style-guide suggestions | United States [Owner: confirm] | Fallback and faster AI model, frame summaries, glossary | |
| OpenAI API | OpenAI | Nothing in normal use. Older clients could still request an OpenAI model by name | United States | Legacy option, listed until it is removed [Owner: remove the legacy model names, then drop this row] |
| LanguageTool public API | LanguageTooler GmbH | Source-language texts of the open project, sent directly from the plugin | Germany [Owner: confirm] | Spell check |
| Firebase Authentication | Google account email, name, user ID, IP address | United States / EU [Owner: confirm] | Sign-in | |
| Google Cloud: Cloud Run, Cloud Tasks, Secret Manager, Cloud Logging, Firebase App Hosting | Every request to our servers; job IDs (Cloud Tasks); our credentials, including encrypted integration tokens (Secret Manager); server logs | Belgium (europe-west1) for production servers; website hosting [Owner: confirm region] | Hosting, background jobs, secrets, logs | |
| Neon Postgres | Neon (on AWS) | Everything we store: accounts, projects, keys, translations, context, usage records | Frankfurt, Germany (AWS eu-central-1) | Database |
| Auth0 | Okta | Email and Google sign-in identity, connector tokens | EU | Sign-in for the MCP connector, only if you connect an AI agent |
| Crowdin | Crowdin | Keys, source texts, translations; your Crowdin token is used to call it | [Owner: confirm] | Sync, only if you connect a Crowdin project |
| Google Analytics | Pages viewed, button clicks, device, browser, approximate location | United States | Website analytics, only after you allow it | |
| Resend | Resend | Contact form messages: your email, name, topic and message | United States [Owner: confirm] | Delivers contact form messages to our inbox |
| Zoho Mail | Zoho | Emails you send us, and our replies | United States (Zoho US data centre) | Our support inbox |
[Owner/lawyer: confirm a data processing agreement is in place with each company, and the transfer mechanism for those outside the EEA.]
How long text is kept
- In our database: your keys, translations, context and glossary stay until you delete them or the project. Point-in-time backups go back 7 days.
- In our logs: server logs are kept for 30 days (the Google Cloud Logging default). Since September 2026 we do not write AI prompts or responses to the logs; errors can still include short messages.
- AI usage records: one row per AI request, with token counts and no text. [Owner: set a retention period.]
- At the AI providers: text is processed to return a result and kept as set by each provider's API terms. [Owner: confirm each provider's current retention and that API data is not used for training, and state it here.]
- At LanguageTool: as set by LanguageTool's terms for its public API. [Owner: confirm, or move to a paid or self-hosted LanguageTool before launch.]
Questions
Write to privacy@mondoword.com.