Subprocessors
The third parties that process your data, what each receives, and why
Running Speakeazy means relying on other companies for hosting, storage, AI analysis, payments, email and error monitoring. Each of them is a “subprocessor”: a third party that processes personal data or your content on our behalf, under contract, and only to deliver the service.
This is every provider we currently use. We publish it because you are trusting us with manuscripts that are often unpublished and under embargo, and you should be able to see exactly who else is in the chain before you upload anything — not after.
| Provider | Purpose | What they receive | Location |
|---|---|---|---|
| Vercel | Application hosting, plus Vercel Analytics and Speed Insights for traffic and performance measurement on our own pages | Everything you send to the service passes through Vercel in transit, including manuscript text, though none of it is stored there. Also IP address, device and browser information, and pages visited | United States |
| Neon | Managed PostgreSQL database | Account details, project metadata, manuscript text, character profiles and annotations | United States (us-east-2) |
| Anthropic | AI analysis — character identification, dialogue attribution, chapter detection | Manuscript text sent for analysis | United States |
| Modal | Compute that runs the analysis pipeline, including the BookNLP stage and the calls to Anthropic | Manuscript text and analysis results. Modal holds database credentials in order to read and write them | United States |
| Cloudinary | Media storage and delivery | Voice recordings, character reference images, profile pictures and project cover images | United States |
| LemonSqueezy | Payment processing and subscription management (merchant of record) | Name, email address, billing address and payment details. We never receive or store your full card number | United States |
| Resend | Transactional email — password resets, analysis notifications, share invitations | Name, email address, and the contents of the message — which includes your project titles, since notification and share emails name the project | United States |
| Sentry | Error monitoring | Error messages and stack traces, and browser information. IP collection is disabled (sendDefaultPii: false), and manuscript text is stripped before events are sent | United States |
- Purpose
- Application hosting, plus Vercel Analytics and Speed Insights for traffic and performance measurement on our own pages
- What they receive
- Everything you send to the service passes through Vercel in transit, including manuscript text, though none of it is stored there. Also IP address, device and browser information, and pages visited
- Location
- United States
- Purpose
- Managed PostgreSQL database
- What they receive
- Account details, project metadata, manuscript text, character profiles and annotations
- Location
- United States (us-east-2)
- Purpose
- AI analysis — character identification, dialogue attribution, chapter detection
- What they receive
- Manuscript text sent for analysis
- Location
- United States
- Purpose
- Compute that runs the analysis pipeline, including the BookNLP stage and the calls to Anthropic
- What they receive
- Manuscript text and analysis results. Modal holds database credentials in order to read and write them
- Location
- United States
- Purpose
- Media storage and delivery
- What they receive
- Voice recordings, character reference images, profile pictures and project cover images
- Location
- United States
- Purpose
- Payment processing and subscription management (merchant of record)
- What they receive
- Name, email address, billing address and payment details. We never receive or store your full card number
- Location
- United States
- Purpose
- Transactional email — password resets, analysis notifications, share invitations
- What they receive
- Name, email address, and the contents of the message — which includes your project titles, since notification and share emails name the project
- Location
- United States
- Purpose
- Error monitoring
- What they receive
- Error messages and stack traces, and browser information. IP collection is disabled (sendDefaultPii: false), and manuscript text is stripped before events are sent
- Location
- United States
What none of them may do
We do not permit any provider on this list to use your content for their own purposes or to train AI models on it, and each is engaged under its standard terms of service for that purpose. For the AI processing specifically, that is a written contractual term rather than a matter of trust: we access Claude through Anthropic's Commercial Terms, which state that “Anthropic may not train models on Customer Content from Services.”
We do not sell your personal information, and we do not share it for advertising.
Changes to this list
We update this page when a provider is added or removed. It reflects our current providers as of the date below rather than a fixed roster, so it is worth re-reading if this matters to you.
Questions
If you need more detail than this page gives — specific regions, retention periods, or a copy of a data processing agreement — ask and we will provide what we can.
Last updated: September 7, 2026