Why I built Rift

Every AI tool I use starts from zero. I would solve something with Claude on Monday, then re-explain the same context to Cursor on Wednesday, paste old threads into ChatGPT, and keep a folder of .md notes I had to remember to attach. The thinking I had already done lived scattered across a dozen apps, and none of them could see the others.

I wanted 1 private memory that every tool could draw from, without shipping my conversations to someone else’s server. Nothing did that, so I built it.

Rift quietly captures your local agent sessions — Claude Code, Codex — as you work, with no exports or copy-paste, and indexes them into 1 private archive on your Mac. Past chats from ChatGPT, Claude, Grok, or Gemini come in too, with a one-time export import. You search the whole thing in a keystroke. And because it speaks MCP, the tools you connect, such as Claude, Claude Code, Codex and Cursor, pull the right context themselves, so the thinking you did in one place shows up in the others and you stop re-explaining what you already solved. Importing your ChatGPT history puts those chats in Rift. It does not let the ChatGPT app search Rift: only a tool that can run a local MCP server can.

It is local-first by default. A fresh install makes 0 AI calls and search runs on keywords alone. A few things can leave, but only ones you switch on, and they go to your own accounts. The full story is on the privacy page.

Rift is in private beta and I am building it in the open. If you have a question, an idea, or you hit a rough edge, reach out on X or LinkedIn. I read everything.