Subprocessors
Every third party that may process customer data on our behalf, and what each one does. This list is maintained against the live model catalog and platform configuration — not curated by hand — so it reflects where work actually goes.
1.AI model providers
When a job needs a model, Accentor routes it to one of these providers and sends the content that job requires. Which one runs a given job depends on the capability and on our routing criteria; the receipt attached to every output records the actual provider used.
| Provider | Purpose | Primary location |
|---|---|---|
| Replicate | Model hosting — image, video, audio and 3D generation | United States |
| fal.ai | Model hosting — image, video and audio generation | United States |
| OpenRouter | Language-model routing across multiple upstream vendors | United States |
| Alibaba Model Studio | Language, image and video models (DashScope) | Singapore / China |
| Together AI | Language and image model inference | United States |
| Cerebras | High-throughput language model inference | United States |
| ElevenLabs | Voice synthesis | United States |
| MusicGPT | Music generation | United States |
| Hugging Face | Model hosting and inference | United States |
2.Infrastructure and operations
These process customer data as an inherent part of running the service.
| Provider | Purpose | Primary location |
|---|---|---|
| Supabase | Managed Postgres, authentication and object storage — the primary store for workspace content | United States (EU/APAC available on Enterprise) |
| Railway | Application hosting and compute for the platform services | United States |
| Cloudflare | DNS and domain provisioning for published sites | Global |
| Vercel | Hosting for sites you publish from Accentor | United States |
| Stripe | Payment processing and usage metering | United States |
| Sentry | Error monitoring and diagnostics | United States |
| Google Analytics 4 | Aggregate product and marketing analytics | United States |
3.Used only when you use the feature
These receive data only as a direct result of an action you take. If you never use the capability, they never see anything.
| Provider | Purpose | Primary location |
|---|---|---|
| GitHub | Source repositories and pull requests — only when you use the code capability or connect a repository | United States |
| Roblox (Open Cloud) | Game publishing — only when you publish a game you built | United States |
| DocuSign | Electronic signature — only when you send a document for signature | United States |
| PandaDoc | Electronic signature — only when you send a document for signature | United States |
4.Using your own provider contracts
Enterprise customers can route model work through their own agreements with providers such as OpenRouter, fal.ai, Together AI, Alibaba Model Studio or Cerebras. In that arrangement the provider is your subprocessor rather than ours, under your contract and your terms, while Accentor continues to plan, gate, retry and receipt the work. See Enterprise.
5.Changes to this list
We add and remove model providers as the model landscape changes — this list moves more often than our other policies.
- The version published here is always current for the deployed product.
- Customers with a signed DPA receive advance notice of a new subprocessor as that agreement provides, and may object on reasonable data-protection grounds.
- To be notified of changes, email [email protected] and ask to be added to the subprocessor notice list.
Locations and transfers
“Primary location” is where the provider principally operates the service we use; several are global and may process data in more than one region. Cross-border transfer mechanisms are addressed in the DPA.
Questions about this document? [email protected]