Subprocessors
Last updated: 5 November 2026
Third parties Pure Reason Inc. (“Kylon”) engages to process customer data, with where each one processes it. This list is Annex III of our Data Processing Agreement. Transfers outside the EEA, UK and Switzerland are covered by the Standard Contractual Clauses in that agreement.
When we update this list to announce an intended addition or replacement of a sub-processor, we will notify each affected customer’s designated contact in writing by appropriate electronic means at the same time, or otherwise notify the registered workspace owner. The notice will identify the sub-processor, its role, processing locations, intended effective date and objection deadline. We will give notice sufficiently in advance to allow a reasonable opportunity to object before the proposed sub-processor begins processing the customer’s personal data, in accordance with any applicable notice period agreed with the customer. Additional contacts can subscribe by writing to privacy@kylon.io. Customers may raise reasonable data-protection objections during the notice period. We work to resolve them before the affected processing begins; if no suitable resolution is available, the customer may terminate the affected service as described in Section 5 of the DPA.
Infrastructure
| Sub-processor | Function | Processing location |
|---|---|---|
| Google Cloud Platform | hosting, storage, compute, primary database (AlloyDB), cache | United States (us-central1) |
| Amazon Web Services | AI model hosting (Bedrock) | United States |
| TiDB Cloud | user-built app databases | United States |
| Databricks (Lakebase) | user-built app databases | United States |
| Clerk | authentication | United States |
| E2B | agent sandbox | United States |
| Vercel | user-built app hosting | United States; static content cached on a global CDN |
| Composio | connectivity for some third-party integrations (not native Google OAuth / direct Google API connections) | United States |
| Cloudflare | CDN, DDoS protection, DNS | Global edge network |
| Doppler | secrets management | United States |
AI & Voice Models
| Sub-processor | Function | Processing location |
|---|---|---|
| Anthropic | AI model provider | United States; inference may run in other regions |
| OpenAI | AI model provider | United States; inference may run in other regions |
| Google (Gemini / Vertex AI) | AI model provider | United States; inference may run in other regions |
| Cerebras | AI model provider | United States |
| OpenRouter | AI model routing | United States; routed to the selected model provider's region |
| Novita AI | AI model routing | United States; routed to the selected model provider's region |
| ElevenLabs | voice synthesis | United States |
| Cartesia | voice synthesis | United States; nearest regional endpoint |
| Deepgram | speech-to-text | United States |
| Fish Audio | speech-to-text | United States and Asia-Pacific |
| Vapi | voice calls | United States |
| Twilio | telephony for voice calls | United States |
| LiveKit | voice/audio call infrastructure | United States; media relayed through the nearest server |
Task context can include Google email, event, file, document, or spreadsheet content when you authorize an Agent to process it. The selected model and route determine which AI services receive that context; listing a provider here does not mean it receives every task. Google user data and its derivatives are subject to the Limited Use, AI-use, and sharing restrictions in our Privacy Policy. We do not use Google Workspace API data or its derivatives to develop, improve, or train general-purpose or non-personalized AI or machine-learning models, and do not permit recipients processing that data for a Kylon task to do so. Terms and settings for Google-data processing must satisfy those restrictions; this list does not certify a provider’s retention configuration or contractual terms.
An Agent brought by a member can run locally or in a separately hosted environment and use that operator’s configured AI services. These task recipients depend on the Agent configuration and are not necessarily represented by Kylon’s own subprocessor list. Review the relevant disclosure and access grants before enabling that Agent to process Google data.
Operations
| Sub-processor | Function | Processing location |
|---|---|---|
| PostHog | product analytics | United States |
| Grafana | logging | United States |
| Resend | United States | |
| Stripe | payments | United States and other Stripe locations |
| Firebase | push notifications (Android) | Global Google infrastructure |
| Apple | push notifications (iOS/macOS) | Global Apple infrastructure |
Services you connect yourself (e.g. Google, Gmail, Slack, GitHub, Notion) are authorized by you under your own account and are not Kylon subprocessors. Native Google connections use Google OAuth and direct Google APIs, separately from Composio. Connecting a Google account is also distinct from the Google Cloud hosting and Google AI processing services listed above. Likewise, third-party tools Kylon can call on your behalf through its integration proxy (e.g. Apollo and similar data services) act only when you direct them and are not Kylon subprocessors. We update this list, with the notice described above, as subprocessors change.
© 2026 Pure Reason Inc. All rights reserved.