MediaWriter is a layer over other people's models and services. This is the full list of who they are, what they do, and what of yours can reach them.
Version 2026-07-25 · Effective 25 July 2026
Draft — not yet reviewed by a lawyer.
Still a placeholder insrc/lib/legal/terms.ts: Legal contact address, Privacy contact address, Copyright agent address. This notice disappears once they are filled in. Do not open public sign-up while it is showing.
Generated from the integrations that actually exist in the product rather than maintained by hand, so it cannot quietly fall out of date. Last reviewed 2026-11-03.
| Service | Purpose | What can reach it |
|---|---|---|
| DigitalOcean | Hosting, and object storage for large promo uploads. | Everything stored in your workspace: uploads, drafts, chats, and account records. |
| Cloudflare | Network edge, bot protection, and access control. | Request metadata and IP address. Not your uploads or drafts. |
| Resend | Transactional email — password reset and address verification. | Your email address and the contents of those messages. |
| Optional sign-in with a Google account. | If you use it: your name, email address, and profile image from Google. |
| Service | Purpose | What can reach it |
|---|---|---|
| Anthropic | Primary model for chat, drafting, and analysis, including reviews of generated ad images. | Prompts, promo text, uploaded documents and images sent for analysis, and generated ad images sent for review. |
| OpenAI | Alternative models, and embeddings for search over your work. | Prompts and the text being embedded or generated. |
| Groq | Speech-to-text transcription of the videos and VSLs you upload. | The audio of the files you ask MediaWriter to transcribe. |
| Google (Gemini and Vision) | Alternative models, and image understanding. | Prompts and images sent for analysis. |
| xAI | Alternative model available in the model picker. | Prompts sent to that model. |
| DeepSeek | Runs the default compliance check on drafts, plans and summarizes project research, writes hook variations, picks up writing preferences from your feedback, and runs the DeepSeek models in the model picker. | Drafts, promo text, research findings and feedback notes sent for those steps, and prompts sent to those models. |
| Alibaba Cloud (Model Studio) | Runs the Qwen, Kimi and MiniMax models in the model picker, through Model Studio (DashScope). | Prompts, text and images sent to those models. |
| OpenRouter | Router used to reach additional models. | Prompts sent to models selected through it, which OpenRouter forwards to the underlying provider. |
| Cohere | Re-ranks search results over your own material. | Search queries and candidate passages from your workspace. |
| Service | Purpose | What can reach it |
|---|---|---|
| fal.ai | Image and video generation. | Generation prompts and any reference images you supply. |
| Runway | Video generation. | Generation prompts and source stills. |
| TopView | Avatar and product video generation. | Scripts, prompts, and source media for the clip. |
| Arcads | Talking-actor video generation. | Scripts and actor selections. |
| HeyGen | Avatar video generation, in the operator's workspace only. Private member workspaces do not use it. | Scripts and avatar selections. |
| ElevenLabs | Text-to-speech voiceover. | Script text to be spoken, and any voice sample you provide. |
| Twelve Labs | Video understanding and scene analysis. | Video files submitted for analysis. |
| Service | Purpose | What can reach it |
|---|---|---|
| Perplexity | Web research used to ground drafts. | Research questions, which may include your product details. |
| Tavily | Web search for research runs. | Search queries. |
| Federal Reserve Bank of St. Louis (FRED) | Optional official economic readings and release dates for enrolled Research projects or an explicit free preview. | Selected public series identifiers; no drafts or uploads. |
| U.S. Bureau of Labor Statistics | Optional public price and employment series. | Selected public series identifiers and date range. |
| Library of Congress / U.S. Government Publishing Office | Bill-title monitoring and optional GovInfo publication-date confirmation. | Date ranges and public bill identifiers; project topic matching runs locally. |
| Federal Register | Public notices and regulatory updates for an explicit Research preview or enrolled project. | Project topic search terms. |
| The GDELT Project | Aggregate topic tone and news volume. | Project topic search terms. |
| Alternative.me | Crypto Fear & Greed Index for crypto projects only. | Request for the public index; no project text. |
| U.S. Securities and Exchange Commission | Optional public filings search for advisor and capital-raise Research projects. | Project topic search terms and the operator's identifying User-Agent. |
| You.com | Web search for research runs. | Search queries. |
| Apify | Collects public competitor advertising you ask us to look at. | The URLs and search terms you submit. |
| Meta Platforms | Only if you connect your Meta ad account: reading your campaigns and ads. | Your Meta access token and the ad-account requests you make. |
| Service | Purpose | What can reach it |
|---|---|---|
| Pexels | Stock footage and photography. | Search terms only. |
| Unsplash | Stock photography. | Search terms only. |
| Coverr | Stock footage. | Search terms only. |
| Storyblocks | Licensed stock footage. | Search terms only. |
| Vecteezy | Licensed stock footage and graphics. | Search terms only. |
Meta Ads is used only when you connect your own Meta ad account, listed above. Google Analytics and Umami connect to the operator’s own accounts and are not reachable from a customer workspace, so no customer content passes through them.
We will update this page before a new subprocessor begins handling customer content. If you need advance notice of changes, write to [[email protected] — mailbox must exist and be monitored].