Engram
Gives coding agents a persistent, searchable project memory stored locally in SQLite
Gives an assistant long-term memory: verbatim conversation storage with semantic search and a knowledge graph
By MemPalace Python Knowledge and memory
67 popularity score ★ 60k
MemPalace is a local-first memory system for AI assistants. It stores conversation history and other content as verbatim text, without summarising or paraphrasing, and retrieves it with semantic search. The index is structured: people and projects become wings, topics become rooms and the original text lives in drawers, so searches can be scoped. The default search backend is ChromaDB, and it also keeps a temporal entity-relationship graph in local SQLite. Its README says nothing leaves your machine unless you opt in.
The MCP server gives an assistant about 45 tools: read tools for status, wings, rooms and semantic search; write tools to file, update, mine and delete drawers; knowledge-graph tools to add, query, invalidate and supersede facts with timelines; navigation across wings; an agent diary; and agent-coordination tools for tasks, events and artifacts. On first use, mempalace_status hands the assistant a memory protocol: search before answering about people or past events, and write diary entries after each session.
Install the PyPI package mempalace (uv tool install mempalace or pipx), then add the server to a client with python -m mempalace.mcp_server, optionally with --palace /path/to/palace; mempalace mcp prints the exact commands for your setup. A Docker image, ghcr.io/mempalace/mempalace, runs the MCP server over stdio with a volume mounted at /data.
mempalace_statusmempalace_searchmempalace_list_wingsmempalace_list_roomsmempalace_get_taxonomymempalace_check_duplicatemempalace_add_drawermempalace_update_drawermempalace_delete_drawermempalace_delete_by_sourcemempalace_minemempalace_checkpointmempalace_kg_querymempalace_kg_addmempalace_kg_invalidatemempalace_kg_timelinemempalace_traversemempalace_diary_writemempalace_diary_readmempalace_task_createmempalace_event_appendmempalace_artifact_put