Skip to main content

​
Persistent memory for coding agents

This is one of the best use cases memloom was built for. Register memloom as an MCP server in Claude Code, Claude Desktop, or Cursor (setup), and the agent gets save_memory / recall_memory tools backed by your store:
  • The agent saves what it learns, like “deploys go through scripts/release.sh, never manual” (procedure) or “user prefers using vue over next.js for frontend development” (preference), and recalls it in later sessions.
  • Because every agent talks to the same daemon, memory saved in Claude Code is there when you switch to Cursor, Codex, or any other MCP-compatible agent/IDE. The memory belongs to you, so you’re not locked into any vendor.
  • Dedup means agents can save enthusiastically without flooding the store, and contradiction detection means a confused agent can’t silently overwrite a valid fact.

​
Personal, unified second brain

Point memloom at your files and they become part of the same memory. If you know NotebookLM: this is that idea, but local, open source, and shared with every agent you connect.
  • Your .md/.txt/.pdf files are recallable by meaning and by exact phrase, with citations back to the file, section, and page.
  • The store is the one place your knowledge lives. Every agent and assistant you connect reads from it, instead of each tool keeping its own partial copy.
  • Edit a file and re-run context add: changed files are re-ingested, unchanged files are skipped (content-hash check), so the store stays in step with your folder.