Honest answers, even where it costs us.
Everything people ask about Scroll Coach, including the questions with unglamorous answers. If yours is missing, ask us directly.
What it is and how it works.
What is Scroll Coach?
A personal library for the content you capture from social media. You share a post into the app, it extracts what was actually said and shown, and it stores that as a private, searchable corpus. Any AI you already use, like Claude or ChatGPT, can then read your corpus through a standard connection called MCP.
How does capturing work?
You share a post from Instagram, TikTok, X, YouTube, or the web using your phone's normal share sheet and pick Scroll Coach. You can chat about the capture right away or send it straight to your corpus. Either way, extraction runs in the background: metadata, transcript, on-screen text, and a structured analysis. You can also paste a link directly in the app.
Which platforms are supported?
Today: Instagram (including reels and carousels), TikTok, X, YouTube, Threads, Facebook, Pinterest, Substack, and general web articles.
Reddit and LinkedIn are not supported: their current terms do not allow the kind of access we would need, and we only build on platforms where we can operate cleanly. If that changes, we will support them properly rather than quietly.
What does the AI actually get out of a video?
For videos the AI watches the content and records a speaker separated transcript, on-screen text, music and notable sounds, and a description of what changes visually over time. YouTube videos with an official caption track keep that word perfect track, with timestamps, instead of a full AI watch. On top of that, every capture gets a clean brief plus a deep structured record: topics, key points, people and products mentioned, claims, quotes, and steps. That depth is what makes a capture genuinely findable later.
How is this different from asking an AI about a link?
Most assistants cannot open social links at all: we verified that ChatGPT, given an Instagram reel link, asks you to upload the video or paste a transcript instead. Scroll Coach captures the content from your own device, where you already have access, extracts it in depth, and keeps it in a corpus you own.
Their AI answers and forgets. Your corpus remembers, and any AI you choose can use it.
Your corpus, your rules.
Is my data private?
Your corpus is yours alone. Every record is isolated per user at the database level, analytics never contain your content, and external AI access is read only. Video files fetched for analysis are deleted the moment analysis finishes: we keep the analysis text, a thumbnail, and metadata, not your media.
The architecture details live on the Your data page.
Does AI ever process my content?
Yes, and we are upfront about it: building the analysis is AI work. Anthropic models produce the written analysis and power chat, and Google Gemini watches videos. Content goes to them only to produce your result, and they do not train on it.
Search embeddings are different: they are computed inside our own database boundary and your text is never sent to an external embeddings service.
Do you scrape social media?
No, never. Nothing is collected unless you personally share it, from your own account, in your own session, on your own device. That is not just policy, it is how the product is built.
Do you store my videos?
No. When a video needs to be analyzed, the media is processed and then deleted the instant analysis finishes. Your corpus stores the understanding of the content: the transcript, the analysis, a thumbnail, and the link back to the original.
Can I delete my data?
Yes, completely. Deleting a capture is a true hard delete on your device and in the cloud, and Delete All My Data removes everything, revoking AI access tokens first so nothing can read your corpus while it is being deleted. You can export your whole corpus as JSON or Markdown at any time before you go.
Connecting the assistants you already use.
What is MCP and what does it mean for me?
MCP is an open standard that lets AI assistants connect to outside data sources. Scroll Coach runs an MCP server for your corpus, so the AI you already use can search and read your captures when you ask it to. The connection is read only: your AI can look things up, but nothing can write to or change your corpus.
Does it work with ChatGPT and Claude?
Yes. The connection is verified working from Claude Desktop and from ChatGPT, including ChatGPT on mobile. In ChatGPT you enable it through Developer Mode, and we will provide a short guide for that.
The unglamorous but honest part.
When can I get it?
Scroll Coach is in private development on Android and is not downloadable yet. A beta comes next. Join the waitlist and we will tell you the moment it opens.
Android or iOS?
Android first. iOS is planned for later, and joining the waitlist is the best way to hear when it lands.
What does it cost?
Pricing is not set yet. We will announce it before the beta opens, and waitlist members hear first. The principles are already fixed and published on the pricing page: a real free tier, fairly priced paid plans, usage always visible, and no lifetime gimmicks.