memcell

docsguidesacross agents

Can my agents share one memory across Claude Code, Cursor and the rest?

Yes — if the memory lives outside any one agent. Per-tool memory silos what a project learned inside whichever tool learned it. memcell wires every agent on the machine to the same record, each in its own dialect: a decision captured from a Claude Code session is recalled in Cursor the next morning.

the silo problem the memory belongs to the project, not the tool that was open

Each tool ships its own memory: a rules file here, a memory feature there, none of them shared. Switch tools — or sit beside a teammate who prefers another one — and the project's learned truth stays behind. Worse, two tools each learn the same lesson separately, and neither hears when the other's version fails.

one record, every dialect fourteen agents, one law

npx memcell connect wires the agents on your machine — each in its own hook shape and config home. Capture reads each agent's best session record; recall speaks each agent's own injection dialect. The record underneath is one:

  • a decision made in Claude Code on Tuesday serves in Cursor on Wednesday
  • an outcome reported from a Codex session moves the same confidence Gemini CLI sees
  • agents with native memory keep it — memcell coexists, never competes for the same file

teams, the same way sharing is a space, not an export

The same law scales past one machine: a shared space serves every member's agents, and what one person's session learns — and what failed for them — reaches the whole team's tools on their next session.

npx memcell connect