Memory & Recall

Status: ✅ Current · Last reviewed: 2026-06-18

Memeri remembers your project so your agents never start from scratch. Memory comes in a few layers, all searchable.


What's stored

  • Decisions, constraints, learnings — distilled knowledge (see Intelligence).
  • Job updates — the running work log, with files changed and commits.
  • Scratchpad notes — quick "noticed in passing" notes agents leave for each other.
  • Conversation context — what was discussed (when sync is on).

Recall — "what do we know about X?"

Any connected agent can pull memory back with two tools:

  • get_project_context({ format: "scoped", query: "auth" }) — the rich frame for entering a topic: pinned items, top decisions, current plan, memory hints. Use it once when starting on something.
  • recall_memory({ query: "why magic links" }) — a lighter, semantic lookup you can call freely mid-task. Returns ranked matches across decisions and notes.

Both run on a local embedding model on Memeri's servers — no third party sees your text, and there's no per-query AI cost.


You can browse it too

The Recall page (and the per-chat Memory drawer in the Terminal) let you read and pin what the project knows. Pin anything you want kept handy.

Rule of thumb: get_project_context sets up the meeting room; recall_memory looks something up without leaving it.