Claude has memory now, but it's a black box. You can't see what it remembers, diff what changed, or revert when it garbles something. And it's personal to one user. If you're working with a team, that context doesn't transfer.

dullnote is a two-way workspace where Claude reads your project files at the start of a session and writes back what happened at the end. Notes, decisions, logs all persistent across sessions via MCP.

The part I'm most proud of is the version control layer. Every edit is tracked with full diffs. You can see exactly what changed, who made the edit (you or Claude), and revert anything. It's the safety net that makes it actually usable, you're not just trusting the AI to overwrite your notes blindly.

Solo founder, built the whole thing with Claude Code. Been using it daily for two months. Free tier available.

How are you all handling persistent context across AI sessions? Especially on teams, curious what's actually working.

https://dullnote.com