Sub-processors
Third-party sub-processors engaged by Wilmo to provide the Service. Changes are notified to customers with 30 days' notice per our DPA.
Services a customer connects to Wilmo itself (for example Shopify, Gmail, Facebook, Instagram, or a carrier) are the customer's own providers, not Wilmo's sub-processors.
Last updated: 2026-11
Correction, 15 October 2026: OpenAI, Anthropic, Mistral AI, xAI, TypeSafe AI, Twilio, Upstash, Firecrawl, Browserless, Linear and Slack were already used to provide the Service but were missing from this list. They are listed now as a correction, not as newly engaged sub-processors.
Clarification, 4 November 2026: Railway's purpose now also names the fixed-IP connections Wilmo makes to systems a customer connects (see IP addresses). Railway is an existing sub-processor in the same region; this is not a newly engaged sub-processor.
Core Infrastructure
| Sub-processor | Purpose | Personal data | Region | Link |
|---|---|---|---|---|
| Supabase | Database, authentication, file storage | All Customer Data stored by the Service | EU (Stockholm, Sweden) | |
| Vercel | Application hosting and deployment | All Customer Data processed by the web application | EU (Stockholm, Sweden) | |
| Railway | Background jobs: AI agent runs, email processing, webhook processing; fixed-IP connections to systems the customer connects | All Customer Data processed by background jobs | EU (Amsterdam, Netherlands) | |
| Postmark | Inbound and outbound email delivery | Email addresses and email content of support tickets | EU (Ireland) | |
| Twilio | Phone channel: calls and call recordings | Phone numbers, call audio and recordings | US | |
| Upstash | API rate limiting | IP addresses and user and team identifiers | EU (Frankfurt, Germany) | |
| Firecrawl | Reading the Customer's public website and web search for the AI agent | Website addresses and search queries, which can contain text from support messages | US | |
| Browserless | Headless browser sign-in for the Talkdesk integration | The Customer's Talkdesk sign-in session | Provider-hosted | |
| Stripe | Subscription billing and payment processing | Billing contacts and payment details of the Customer | EU/EEA | |
| Infisical | Secret management | None (application secrets only) | EU | |
| HubSpot | CRM, onboarding, marketing communications | The Customer's own contacts and users, not end customers | EU (Frankfurt) |
AI Inference Providers
AI sub-processors used for model inference. None of them use Customer Data to train models. Every AI model request Wilmo sends through Vercel AI Gateway requires zero data retention, so it is served only by a provider that deletes prompts and outputs after processing. Two features are outside that rule, as set out in DPA section 19.4: phone-call transcription, which sends call audio directly to OpenAI, where it may be kept for up to 30 days for abuse monitoring; and spoken input to Wilbot, which is transcribed by OpenAI through the AI Gateway without a zero-data-retention requirement.
| Sub-processor | Purpose | Personal data | Region | Link |
|---|---|---|---|---|
| Vercel (AI Gateway) | Routes every AI model request to a model provider and requires zero data retention on each request | Prompts and outputs: ticket content and data the AI agent looked up | Global | |
| OpenAI | GPT models and embeddings through the AI Gateway; phone-call transcription called directly (see note above) | Ticket content; phone-call audio | US | |
| Microsoft Azure | OpenAI models through the AI Gateway | Ticket content | Routed by the AI Gateway | |
| Anthropic | Claude models through the AI Gateway | Ticket content and Wilbot conversations | US | |
| AWS (Amazon Bedrock) | Claude models through the AI Gateway | Ticket content and Wilbot conversations | Routed by the AI Gateway | |
| Google Cloud (Vertex AI) | Models through the AI Gateway | Ticket content | Routed by the AI Gateway | |
| Mistral AI | Grammar and clarity fixes of draft replies through the AI Gateway | Draft reply text | Routed by the AI Gateway | |
| xAI | Wilbot chat titles through the AI Gateway | The first message of a Wilbot chat | US | |
| TypeSafe AI (Jev) | Short classification checks through the AI Gateway (Wilbot routing, chatbot name check) | Wilbot messages and chatbot display names | Routed by the AI Gateway |
Monitoring & Analytics
| Sub-processor | Purpose | Personal data | Region | Link |
|---|---|---|---|---|
| PostHog | Product analytics, feature flags, session replay (all text, inputs and media masked), and usage metrics of AI requests | Merchant user identifiers and usage; AI request metrics (model, tokens, latency) without prompts or outputs | EU | |
| Sentry | Error tracking, performance traces, and session replay on errors (text masked, media blocked) | Error context; Wilmo scrubs personal data before sending, but some can remain | US | |
| Better Stack | Application logs, uptime monitoring, distributed tracing, status page | Log lines and traces; Wilmo redacts personal data from logs and keeps AI prompts and outputs out of traces, but some personal data can remain | EU |
Support & Internal Operations
| Sub-processor | Purpose | Personal data | Region | Link |
|---|---|---|---|---|
| Linear | Bug reports and feedback the Customer's users send from the app | Reporter name and email, team, page address, ticket number, and the report text | US | |
| Slack | Internal notifications to the Wilmo team (signups, integration events, bug reports) | Merchant user email addresses and account events; bug-report emails are redacted | US |
Stay informed
We notify all customers by email when this list changes, with at least 30 days' notice before a new sub-processor begins processing. If you have questions or concerns, contact us at privacy@wilmo.ai.