Sub-processors
Last updated: 20 September 2026
Recalld (Bit Robotics Ltd) uses the third parties below to provide the service. Each one processes data only on our instructions, under a data processing agreement. This page is the authoritative, versioned list; we update it and notify customers by email before a new sub-processor starts processing customer data.
Infrastructure and billing
| Sub-processor | Purpose | Processing location |
|---|---|---|
| OVHcloud | Servers and databases for each region: API, memory storage, embeddings, logs | EU region: France (Gravelines). US region: United States (Virginia) |
| Google Cloud (Cloud Storage) | Encrypted database backups for each region, kept for 14 days | EU region: europe-west1 (Belgium). US region: us-east4 (Virginia) |
| Cloudflare | Marketing website hosting, DNS and cookie-free visit statistics for the marketing website. API and dashboard traffic goes straight to our servers and does not pass through Cloudflare | Global network (EU/US) |
| Stripe | Payment processing, subscriptions and invoicing | EU / US |
Model providers
These providers run the language-model and embedding calls used to extract facts from ingested content, answer recall requests and power the chat application. Which providers are involved depends on the models configured for your account and the models you select in chat. None of them may use your data to train models.
| Sub-processor | Purpose | Processing location |
|---|---|---|
| Google (Vertex AI and Gemini API) | LLM inference and embeddings for memory extraction and recall; chat models | European Union when served through Vertex AI regional endpoints; otherwise Google's global infrastructure, which may include the United States |
| Microsoft (Azure OpenAI Service) | LLM inference and embeddings | Depends on the configured endpoint (EU or US) |
| OpenAI | LLM inference (chat models) | United States |
| Anthropic | LLM inference (chat models) | United States |
| Cerebras | LLM inference (chat models) | United States |
| DeepInfra | LLM inference and embeddings for open-weight models (chat models) | United States |
| OpenRouter | Routing to open-weight chat models hosted by its own providers | United States |
| Tavily | Web search for the chat application, only when you enable search in a chat. Receives the search query, not your stored memory | United States |
For providers processing data outside the UK/EEA, transfers are protected by the provider's data processing terms incorporating the EU Standard Contractual Clauses with the UK Addendum, or, where the provider is certified, the EU-US Data Privacy Framework and its UK Extension.
Bring your own key
If you connect your own model provider key (BYOK), calls made with that key run under your own agreement with that provider. In that case the provider acts for you directly and is not a Recalld sub-processor.
Questions and the DPA
This list forms part of our Data Processing Agreement. To receive change notifications for this list, or ask anything else, email support@recalld.ai.