Subprocessors

1. How to read this register

This register identifies material third-party providers evidenced in the current Rainspeed Labs service architecture. Certain provider names may be omitted from this public register while their processing category and purpose remain described. The provider roster may change and is maintained in our internal records. A provider is used only for the relevant product, feature, deployment or account configuration. Listing a provider does not mean that every request or Service uses it.

The register separates providers used by Rainspeed Labs to operate a Service from direct providers that a user chooses and configures in ONeNas Code. A model selected through a routing provider may also be operated by another model developer.

Providers may be added, replaced or removed as the Services evolve. We will update this register when a material provider changes and provide notice required by an applicable agreement or law.

2. Identity, authentication and anti-abuse

• Clerk - account authentication, identity, sessions and related security events across applicable hosted AtlasFlux and Support services, including selected Google or GitHub sign-in flows through Clerk.
• Cloudflare Turnstile - anti-bot verification where enabled on AtlasFlux AI authentication or other protected flows.

3. Hosting, databases and storage

• Vercel - hosting and serverless deployment for applicable web applications and services.
• Fly.io - deployment and runtime infrastructure for applicable AtlasFlux API, agent, media and ONeNas relay services. The current AtlasFlux API deployment uses Singapore as its primary region.
• Supabase and PostgreSQL - database and, for applicable AtlasFlux AI features, storage infrastructure for accounts, chat, media, credits, community and operational records. The API project can use Supabase-hosted PostgreSQL through its DATABASE_URL, while the Support service also performs configured read-only lookups against product Supabase projects.
• Neon - PostgreSQL database infrastructure for AtlasFlux Support.
• Cloudflare R2 - object storage for applicable AtlasFlux AI generated files, exports, agent artifacts and public media, and for private AtlasFlux Support attachments. AtlasFlux AI and Support use different access patterns; Support attachments use signed URLs, while some AtlasFlux AI objects are public.

4. Payments and communications

• Stripe - checkout, payment processing, payment methods, subscriptions or automatic reloads where enabled, refunds, chargebacks, receipts and payment webhooks.
• Resend - transactional, account, newsletter and Support email delivery where enabled. Support stores an email outbox and delivery events in its database.
• Svix-compatible webhook verification - the Support application uses the Standard Webhooks/Svix signing scheme and library to verify Resend webhook requests. Rainspeed Labs does not use an independent Svix database or content store for Support messages.

5. AI, search and media processing

• Anthropic - direct model inference for configured AtlasFlux AI or agent operations and for direct provider configurations in ONeNas Code.
• ILMU - configured AtlasFlux AI text-to-speech or voice operations, and direct ONeNas provider configurations where enabled.
• OpenAI - configured AtlasFlux AI tool or media operations where enabled, and direct ONeNas provider configurations where selected by the user.
• Exa - web search, retrieval and content extraction where enabled.
• Tavily - fallback or configured web search where enabled.
• ScrapingFish and scrape.do - retrieval or rendering of selected public web pages where enabled.

6. Agent tools and sandboxed workflows

• E2B - temporary sandboxed execution and file-generation environments for applicable AtlasFlux agent and document workflows.
• Context7 / Upstash Context7 MCP - documentation retrieval and MCP tool operations for applicable AtlasFlux agent workflows.

7. Direct providers selected by the user

ONeNas Code supports a managed AtlasFlux provider and can also expose direct provider configurations, depending on the build and configuration. Examples evidenced in the ONeNas documentation include OpenAI, Anthropic, Google, a model-routing provider, xAI, Groq, Mistral, NVIDIA, GitHub Copilot and ILMU, as well as custom OpenAI-compatible endpoints.

When you configure a direct provider, your credentials, prompts, selected context and Output may be sent directly to that provider. These providers are not necessarily Rainspeed Labs subprocessors, and their handling is governed by their own terms and privacy policies. Review the provider before enabling it and do not assume that managed AtlasFlux retention or training terms apply.

8. Changes and questions

Questions about a provider register or the company arrangements behind a feature may be sent to hello@rainspeedlabs.com. Product-specific support questions should be sent through https://support.atlasflux.my.