Subprocessors
Last updated: September 2, 2026
These are the companies that process data on our behalf when you use Feerasta. We publish the list rather than holding it behind a request, because a subprocessor list that takes a week to obtain adds a week to every legal review.
We update this page before a new provider processes customer data. The list is reconciled to what the code calls today, not to what a roadmap names. Inside the product, a provider disclosure shows which of these your own workspace actually uses. Our data processing agreement is a template signed on request at hello@feerasta.ai.
Infrastructure
| Provider | What it does for us | Where it processes |
|---|---|---|
| Hetzner | The server running the Feerasta platform (app.feerasta.ai) and its PostgreSQL database | Germany (Falkenstein) |
| Cloudflare | Website delivery, DNS, edge compute and DDoS protection | Global edge network |
| Hostinger | Domain, DNS and mail hosting for our own mailboxes | European Union |
Communication
| Provider | What it does for us | Where it processes |
|---|---|---|
| Twilio | Phone numbers, calls and SMS | United States |
| ElevenLabs | Speech synthesis and recognition for voice agents | United States |
| Meta | The WhatsApp Business API: WhatsApp messages and phone approvals | United States |
| Resend | Transactional email delivery (welcome, login, reset, invite, receipt), on Amazon SES infrastructure | United States |
AI model providers
Customer-facing conversations on the web widget, SMS, WhatsApp and voice are answered from your approved rules through one resolver and make no model call; a conversation is never sent to a model provider. The providers below are used only for model-backed drafting that waits for your approval, and only where you have not brought a model key of your own. Where you bring a key, that provider replaces these for your workspace and its terms govern those calls.
| Provider | What it does for us | Where it processes | Training on customer data |
|---|---|---|---|
| Anthropic | Language model behind drafting and classification when the platform supplies the model | United States | Not used for training |
| OpenAI | Language model behind a workspace's agent container when the platform supplies the model | United States | See provider terms |
| NVIDIA | Vision model that reads receipts and documents for the WhatsApp Assistant demo (whatsapp.feerasta.ai) only; not used by the platform | United States | See provider terms |
Where a client engages Feerasta Sovereign, models run on infrastructure the client controls and these providers are not used for that client's data.
Business operations
| Provider | What it does for us | Where it processes |
|---|---|---|
| Stripe | Payments and billing | United States |
| Business data, maps and places information; the encrypted offsite copy of nightly backups (encrypted before it leaves our server, so Google cannot read it) | United States |
Not on this list, on purpose
Zapier, Make and similar automation tools are not our subprocessors. If you forward email or invoices to your workspace through one of them, that is your own account and your own terms. No other model provider, analytics vendor or data broker processes customer data.
Being told when this changes
We update this page before a new provider processes customer data. Email hello@feerasta.ai and ask to be notified of subprocessor changes, and we will write to you before a new subprocessor that handles personal data starts processing.
Reporting a security problem
If you have found a vulnerability, our contact details and what to expect are at /.well-known/security.txt. We would rather hear it from you than from someone else.