In short
- Pasting a Reel or TikTok link into ChatGPT or Claude rarely works. The page needs a login, and the assistant can't reliably watch the video.
- The fix is to turn the video into context first: speech, on-screen text, visuals, caption and links, then give that to the assistant.
- Three ways to do it: paste a transcript yourself, hand off a project from ORANO as a single link, or connect ORANO MCP so a compatible agent can search your whole library.
- ORANO MCP is read-only and personal: agents can read your projects, sources, tasks and chosen memory, but can't change anything.
You saved a Reel about a pricing strategy and want ChatGPT to apply it to your business. Or you saved a TikTok listing five developer tools and want Cursor to set one up. You paste the link and the assistant says it can't access it, or it confidently describes a video it never saw.
The problem isn't the assistant. It's the format. Here's why links fail, and three ways to give an AI the real content of the videos you've saved.
Why pasting a video link doesn't work
- Login walls. Instagram and TikTok pages often need a signed-in session, which an assistant's browsing tool doesn't have.
- Video isn't text. Even when the page loads, the useful part is in the speech, the on-screen text and the visuals, not the HTML.
- Every question starts over. Re-processing raw video for each follow-up is slow and expensive, even for tools that can do it.
The fix is the same in every case: process the video once into clean context, then give the assistant that context.
What good video context looks like
An assistant needs more than a transcript. Useful video context keeps these parts separate, so the model knows where each fact came from:
- Speech: what was said.
- On-screen text: lists, names, prices and settings shown in the frames.
- Visual analysis: what the video actually shows.
- Caption and links: the creator's text and any resources mentioned.
- A summary and tasks: the useful result, with steps when the video is actionable.
Keeping the creator's caption apart from the speech, and both apart from outside research, stops the assistant from treating a caption's hype as something the video showed.
Method 1: Paste a transcript (free, manual)
Get the words out of the video. For YouTube, use Show transcript. For Reels and TikToks, see how to get a Reel transcript. Then paste it with a clear instruction: "Here's the transcript of a video I saved. Pull out the method and adapt it to [my situation]."
This works for talking-head videos. It fails when the content is on screen, and it doesn't scale beyond one video at a time.
Method 2: Hand off a project from ORANO (one link)
When you share a video to ORANO, it builds a project from the speech, visuals, on-screen text and source. In the project, tap Connect your agent. ORANO packages the summary, tasks and research as a single link that ChatGPT, Claude, Cursor or a local model like Ollama can read.
This is the quickest way to say "take this saved video and keep going". Use it for a planning conversation in ChatGPT, a writing session in Claude, or a build in Cursor.
Method 3: Connect ORANO MCP (your whole library)
The Model Context Protocol (MCP) is an open standard for connecting AI apps to outside tools and data. ORANO MCP gives a compatible agent read-only access to your saved context, so you don't paste anything at all. The agent can:
- list and open your projects, including their summaries, sources, research and ordered tasks;
- search your library by title, summary, source title or URL ("find the Reel about cold email subject lines");
- read memory facts you've chosen to share, such as preferences or goals;
- pick up handoffs you've sent from the app.
It's read-only: an agent can't change your projects, memory or account. Access uses a personal key that you can rotate or revoke, and the key manager shows which client connected.
Compatibility, plainly stated: ORANO MCP currently needs a client or agent API that accepts a custom Authorization header. Many developer tools and agent frameworks do. One-click OAuth connection isn't available yet, so hosted chat apps that only support OAuth can't connect directly. For those, use Method 2. The agents page and the machine-readable mcp.json have the current details.
ORANO MCP is a $5/month web plan, and you can buy it without installing the app. To see the kind of context an agent receives before you buy, paste a public Reel into the video lab.
Which method should you use?
| Situation | Use |
|---|---|
| One talking-head video, right now | Paste the transcript |
| One saved video you want to act on with AI | ORANO project handoff |
| An agent that should know everything you've saved | ORANO MCP |
| Videos whose content is mostly on-screen text | ORANO (handoff or MCP), because it reads the frames |
Prompts that work well with saved-video context
- "Search my ORANO library for anything about onboarding emails and combine the advice into one checklist."
- "Take the tasks from my saved workout project and fit them into a three-day week."
- "Which of the tools in my saved Reels have a free tier? Cite which video each one came from."
- "Using the Reel I handed off, draft a version of this strategy for a two-person bakery."
Frequently asked questions
Can ChatGPT watch an Instagram Reel?
Not reliably from a link. Instagram and TikTok pages often need a login, and a chat assistant may not be able to fetch or process the video itself. Give it the content as text, or use a tool that processes the video and passes the context along.
What is MCP?
The Model Context Protocol is an open standard that lets AI apps connect to outside tools and data in a consistent way. A compatible client, such as an agent or code editor, can call an MCP server's tools to read context it couldn't reach on its own.
What can an AI agent read through ORANO MCP?
Your ORANO projects, their summaries, sources, research and ordered tasks, a search over your saved library, and the memory facts you've chosen to share. It's read-only, so agents can't edit your projects, memory or account.
Which AI clients work with ORANO MCP?
Any MCP client or agent API that lets you set a custom Authorization header can use your personal key. One-click OAuth connection isn't available yet, so some hosted chat apps that only support OAuth can't connect directly. The in-app project handoff link works with ChatGPT, Claude, Cursor and local models.
How much does ORANO MCP cost?
ORANO MCP is a separate web plan at $5 a month, bought on the pricing page. The ORANO app itself is free to download, with optional ORANO Pro subscriptions in the App Store and Google Play.