Sub-processors
Third-party providers that process data on behalf of Fruxon, and the AI providers and messaging channels you can connect with your own accounts.
Last updated: October 2026
1. Overview
This page lists two distinct categories. Sub-processors (sections 2–3) are vendors Fruxon engages under contract to process data on our behalf; they are covered by the commitments in our Data Processing Addendum (DPA), including advance notice of changes. Customer-connected providers (sections 4–5) are AI providers and messaging channels you connect using your own API keys, accounts or credentials; you engage them directly under your own agreements, and they are not Fruxon sub-processors. This page is referenced by the Terms of Service and the DPA.
2. Infrastructure Sub-processors
Core infrastructure services used to host and operate the platform.
| Sub-processor | Purpose | Data Processing Location |
|---|---|---|
Google Cloud Platform (Google LLC) | Cloud hosting, compute, storage, KMS, Pub/Sub, Secret Manager | Workspace's residency cell: United States (us-central1) or Israel (me-west1) |
Cloud SQL for PostgreSQL (Google LLC) | Primary relational database | Same region as the workspace's residency cell |
Google Cloud — Vertex AI and Cloud Vision (Google LLC) | AI processing on Fruxon's own Google Cloud project: Gemini models offered as “Managed by Fruxon”, image description and text extraction (OCR), and built-in assistants such as version notes and skill generation | Global endpoints — may be served from any Google Cloud region |
3. Service Sub-processors
Additional third-party services used for specific platform functions.
| Sub-processor | Purpose | Data Processing Location |
|---|---|---|
Firebase and Identity Platform (Google LLC) | User sign-in and web app hosting | United States |
Google Workspace — Gmail API (Google LLC) | Transactional and notification email | Global (Google data centers) |
Stripe, Inc. | Payment processing and billing | United States |
Mixpanel, Inc. | Product analytics | United States |
Sentry (Functional Software, Inc.) | Error monitoring for the web application | United States |
Loops | Account sign-up and onboarding email | United States |
Anthropic, PBC | Copilot, the in-product assistant that helps you build agents (runs on a Fruxon-held key; United States cell only) | United States |
OpenAI, L.L.C. | Sandbox test-data generation for simulated integrations (runs on a Fruxon-held key) | United States |
4. Customer-Connected AI Providers
Your agents can call LLM providers using API keys or cloud credentials you supply, under your own agreement with each provider (bring-your-own-key, BYO-Key). Data is sent to a provider only when you configure it and initiate agent executions. Because you engage these providers directly, they are not Fruxon sub-processors — each provider's own terms govern your data once it reaches them. This list is provided for transparency:
| Provider | Purpose | Data Processing Location |
|---|---|---|
OpenAI (OpenAI, L.P.) | LLM inference | United States |
Anthropic (Anthropic, PBC) | LLM inference | United States |
Google — Gemini (Alphabet Inc.) | LLM inference | United States |
Google — Vertex AI (Alphabet Inc.) | LLM inference | Region configurable by customer |
DeepSeek (DeepSeek AI, Hangzhou) | LLM inference | China |
xAI — Grok | LLM inference | United States |
Amazon Web Services — Bedrock | LLM inference | Region configurable by customer |
VoyageAI | Embedding / LLM inference | United States |
Hugging Face | LLM inference | United States / EU |
OpenAI-compatible custom endpoints | LLM inference (customer-specified) | Customer-determined |
Provider data-use practices differ and may change. Some providers may use API inputs and outputs to train or improve their models, while others offer no-training terms. If your requirements prohibit training on your data, review the provider's terms before enabling it.
Gemini models offered as “Managed by Fruxon” are different: they need no key of your own and run on Fruxon's Google Cloud project under our agreement with Google, so they are covered by the Vertex AI row in section 2, not by this list.
DeepSeek — China data processing and model training
DeepSeek processes data on servers located in China, and its terms permit the use of API inputs and outputs to train and improve its models, with no opt-out available. If you configure DeepSeek as a provider, prompts and agent data are transmitted to and processed in China and may be used for training. This may have implications under GDPR, data residency requirements, and other data protection regulations. You are responsible for ensuring compliance with applicable data transfer requirements before enabling DeepSeek.
5. Customer-Connected Messaging Channels
Channels connect your agents to the messaging platforms your organization already uses. Each channel runs on your own account, bot, phone number or app registration with that platform. Where you install Fruxon's published Slack or Google Chat app instead of your own, the messages still live in your workspace on that platform, under your agreement with it. Because you engage these platforms directly, they are not Fruxon sub-processors. This list is provided for transparency:
| Platform | Purpose | Data Processing Location |
|---|---|---|
Slack | Messaging channel | Per your account with the provider |
Microsoft Teams | Messaging channel (your own Azure Bot registration) | Per your account with the provider |
Google Chat | Messaging channel | Per your account with the provider |
Telegram | Messaging channel (your own bot) | Per your account with the provider |
Discord | Slash commands and interactions (your own Discord application) | Per your account with the provider |
Twilio | SMS and WhatsApp (your own Twilio account and number) | Per your account with the provider |
360dialog | WhatsApp (your own 360dialog account) | Per your account with the provider |
TexterChat | WhatsApp (your own TexterChat account) | Per your account with the provider |
The same applies to the other services you connect your agents to — for example CRM, ticketing, storage, databases and web-crawling services. They are accessed with your credentials, under your own agreements, and are not Fruxon sub-processors.
6. Change Notification
We update this page when we add or remove sub-processors. If you have a DPA with us, we will notify you of material changes to our sub-processors (sections 2–3) at least 30 days in advance via email. The customer-connected AI provider and messaging channel lists are maintained for transparency and updated as provider support changes. Subscribe to updates by emailing privacy@fruxon.com.
7. Contact
Questions about our sub-processors? Contact us at privacy@fruxon.com.