Memento: persistent memory for your Cursor agent, hosted over MCP

Showcase Built for Cursor
,
1

Memento is hosted memory for Cursor agents, connected over MCP with a single config entry.

What it does:

  • Persistent memory across sessions, projects, and machines. As you work, it saves conventions, architecture notes, decisions, and gotchas, then recalls the right ones automatically next session.
  • Every memory is a plain markdown page you can browse and edit on the web. No black box: read what your AI thinks it knows and correct it.
  • Personal by default, or create a team org so every teammate’s agent shares the same codebase knowledge. A new developer’s agent starts day one already knowing the architecture.
  • Commands like /remember and /recall when you want to drive it directly.

Free tier available: Memento: Now your AI remembers.

Feedback welcome, especially from folks working in large existing codebases.

1 Like
Grounder: session memory shared between Cursor and Claude Code
5

Being able to read and modify what the agent has stored is a good idea. That’s usually the part you don’t see.

Two things I couldn’t tell. The site says the right memories surface automatically before the AI answers. That’s the part I’m curious about: an MCP tool runs when the model calls it, so for that to happen first, is there an always-apply rule telling it to recall, or something else? And if I stopped using it, could I export the memories as markdown?

I posted a local-files version of this here a couple of days ago, so I’m asking as someone who made different calls and is curious about yours