When you run an app or agent on inference shell, your input is processed by a third-party AI provider. Which provider handles your data depends on which app you choose to run.
This page lists all current upstream providers. It is updated as providers are added or removed. Registered users are notified of material changes at least 30 days in advance per our privacy policy.
AI Model Providers
Text and Chat
| Provider | Models | Data processed |
|---|---|---|
| Anthropic | Claude 4.6, Claude 4.5, Claude Haiku | text prompts, conversation history |
| OpenAI | GPT Image, ChatGPT | text prompts, images |
| Gemini Pro, Gemini Flash, Veo, Imagen | text, images, video | |
| OpenRouter | 25+ LLMs (aggregator) | text prompts |
| Bytedance | Doubao, Seedance | text, images, video |
| Opper | Chat models | text prompts |
| XAI | Grok | text, images, video, audio |
Image Generation
| Provider | Models | Data processed |
|---|---|---|
| FAL AI | FLUX, Stable Diffusion, ControlNet, 40+ models | text prompts, reference images |
| Pruna | Optimized image/video models | text prompts, reference images |
| Bria | Image generation, editing, background removal | text prompts, images |
| Reve | Image generation | text prompts |
| Microsoft | Designer | text prompts |
| Alibaba | Wanx | text prompts, images |
| Phota | Image processing | images |
Video Generation
| Provider | Models | Data processed |
|---|---|---|
| Kling AI | Kling 2.0, image-to-video, lip sync | text prompts, images, audio |
| Topaz | Video upscaling, enhancement | video files |
| Veed | Video processing | video files |
| HeyGen | Avatar video, translation | text, images, audio |
Audio and Voice
| Provider | Models | Data processed |
|---|---|---|
| ElevenLabs | TTS, voice cloning, sound effects, music | text, audio files |
| Inworld | Voice synthesis | text prompts |
Search and Knowledge
| Provider | Models | Data processed |
|---|---|---|
| You.com | Web search, RAG | search queries |
| Tavily | Web search, extraction | search queries |
| Exa | Neural search | search queries |
| ArXiv | Paper search | search queries |
Social and Communication
| Provider | Services | Data processed |
|---|---|---|
| X (Twitter) | Post, read, engage | text, images |
| Discord | Send messages | text |
Infrastructure Providers
These providers support platform operations (not model inference):
| Provider | Purpose | Data processed | Location |
|---|---|---|---|
| Hetzner | primary compute, self-hosted databases (PostgreSQL, TimescaleDB), Redis, Meilisearch | all platform data | Germany |
| Google Cloud Platform | EU replica compute (Frankfurt) | platform data (EU replica) | EU (Germany) |
| Cloudflare | object storage (R2), CDN (cdnjs wordmark font), CAPTCHA (Turnstile) | files, log archives, CAPTCHA tokens, IP addresses (cdnjs font requests) | EU (Western Europe) |
| Stripe | payment processing | billing data (card data never touches our servers) | US |
| Amazon SES | transactional email (via SMTP) | email addresses, notification content | US |
| Simple Analytics | website analytics | anonymized page views (no personal data) | EU (Netherlands) |
| PostHog | product analytics and error tracking | user identifiers, feature usage events, error data | US |
The web app bundles its UI fonts (Mona Sans, Haskoy, Fira Code) with the application. Only the Hack wordmark font is loaded from cdnjs (Cloudflare) at runtime.
Provider changes
We may add or remove providers as the platform evolves. Registered users will be notified of material changes at least 30 days in advance via email.
If you have questions about which provider processes your data for a specific app, contact [email protected].
Last updated: 8 October 2026