List of Sub-Processors

Last updated 12 October 2026

Chat Thing engages with a number of Sub-Processors to support our infrastructure and the normal running of the platform. By agreeing with our DPA, you agree all of these Sub-Processors may have access to Customer Data.

Third Party Sub-ProcessorPurposeApplicable ServicePlace of processing
Supabase Inc - DPAHosting & InfrastructureUsed as primary database storage, file storage and authentication systems.United States
RailwayHosting & InfrastructureOn demand cloud computing and APIsUnited States
CloudflareContent Delivery NetworkWeb infrastructure and web site security including domain name server services, DDoS protection, content delivery network, and SSL certificatesUnited States
OpenRouterAI APIsRoutes bot messages to the model provider you select (OpenAI, Anthropic, Google and others). Also works out which sources an answer used, where a bot shows sources (see below).United States
Open AIAI APIsSpeech-to-text and text-to-speech for the web chat's voice features. Only used where a bot has voice enabled.United States
NylasEmail ConnectivityOptional. Only where you connect an email account to a bot: sends and receives mail on that account's behalf. Not used unless you connect one.United States
StripePayment ProcessorCollection and processing of customer payments and subscriptionsUnited States
PlausibleWeb AnalyticsWeb analytics and usage statisticsGermany
PosthogWeb analyticsWeb analytics and usage statistics and A/B testingGermany
ConvertKitMarketing EmailsEmail notifications and marketing emails testingUnited States

How bot messages reach AI providers

Bot messages reach the underlying AI models through OpenRouter, which routes each request to the model provider you have selected for that bot. Where a bot has voice enabled, audio is sent to OpenAI's API directly for speech-to-text, and reply text is sent for text-to-speech; that path does not go through OpenRouter.

How a bot works out which sources it used

Where a bot has Include sources for responses turned on, Chat Thing shows only the sources an answer actually used. To work that out, once the bot has written its answer we send the following to Jev, a model from TypeSafe, through OpenRouter:

  • the customer's message
  • the bot's answer
  • the knowledge-base passages retrieved for that answer
  • the results of any power-ups the bot used in that answer

This runs on Chat Thing's own OpenRouter account, including for teams that use their own AI provider key. It applies to answers in the web chat, streamed replies from the API channel, and MCP. It is not used for Slack, Discord, WhatsApp, Telegram, email or phone conversations. Bots with Include sources for responses turned off send nothing to Jev.

Each provider handles data under its own terms, linked in the table above. See our Privacy Policy for how Chat Thing handles Customer Data.