Your AI cannot use what you read, watch, or note down unless you paste it in. Sensefold saves articles, threads, YouTube (with transcripts), PDFs (OCR), your own notes, and your ChatGPT / Claude / Gemini / Grok chats in one click, as clean Markdown. Then Claude, ChatGPT, Cursor, Claude Code, or Codex read and write that library over MCP: one address, OAuth, permission tiers, every edit versioned and revertible, Markdown export. No chat of its own. Saving is free; AI runs on credits.
Ask Claude or Cursor about a video you watched last week and watch what happens: it opens a browser, scrapes the page, hunts for a transcript, maybe writes a script, and still ends up with a worse copy than the one you had on screen. That is where most of the friction in working with an AI lives, and it pulls both of you away from the idea itself.
Sensefold is the fix on both ends, and it has no chat of its own.
Capture is one click. The Chrome extension and the iPhone share sheet save what you read (articles, PDFs with OCR, X and Reddit threads), what you watch (YouTube with transcripts), and what you already discussed (ChatGPT, Claude, Gemini, and Grok conversations). A passing thought goes in the same way, on its own or attached to any of them, and the notes you write are never rewritten by AI. Everything lands as clean Markdown, summarized, tagged, and searchable in any language. Saving a link or a half-formed idea is easy enough that you actually do it, so the good stuff stops getting lost.
Then the AI you already use reads it over MCP. Paste one address into Claude, ChatGPT, Cursor, Claude Code, or Codex, sign in, and it can search your library, read the full text and your notes, and write notes back. No browser to drive, nothing to scrape. You set what it may do when you approve the connection (read only, edit, or full), every change it makes is versioned so you can undo it, and the whole library exports as Markdown any time.
Yes, ChatGPT has memory. It lives inside ChatGPT. This is yours, in plain Markdown, and every model you use reads the same copy.
Saving is free; AI processing runs on credits, with a 14-day trial. What would you want your AI to know about what you read, watch, and think? I will be here all day.
Hi Product Hunt, Miles here, maker of Sensefold.
Ask Claude or Cursor about a video you watched last week and watch what happens: it opens a browser, scrapes the page, hunts for a transcript, maybe writes a script, and still ends up with a worse copy than the one you had on screen. That is where most of the friction in working with an AI lives, and it pulls both of you away from the idea itself.
Sensefold is the fix on both ends, and it has no chat of its own.
Capture is one click. The Chrome extension and the iPhone share sheet save what you read (articles, PDFs with OCR, X and Reddit threads), what you watch (YouTube with transcripts), and what you already discussed (ChatGPT, Claude, Gemini, and Grok conversations). A passing thought goes in the same way, on its own or attached to any of them, and the notes you write are never rewritten by AI. Everything lands as clean Markdown, summarized, tagged, and searchable in any language. Saving a link or a half-formed idea is easy enough that you actually do it, so the good stuff stops getting lost.
Then the AI you already use reads it over MCP. Paste one address into Claude, ChatGPT, Cursor, Claude Code, or Codex, sign in, and it can search your library, read the full text and your notes, and write notes back. No browser to drive, nothing to scrape. You set what it may do when you approve the connection (read only, edit, or full), every change it makes is versioned so you can undo it, and the whole library exports as Markdown any time.
Yes, ChatGPT has memory. It lives inside ChatGPT. This is yours, in plain Markdown, and every model you use reads the same copy.
Saving is free; AI processing runs on credits, with a 14-day trial. What would you want your AI to know about what you read, watch, and think? I will be here all day.