mem0ai/mem0's Claude Code Plugin for Persistent Memory

What is mem0ai/mem0?

mem0ai/mem0 is the repository behind Mem0 (pronounced "mem-zero"), a memory layer that gives AI assistants and agents persistent, personalized context across sessions. Flowy lists it under the data category, and at the last crawl the repository carried 64,021 GitHub stars. The Claude Code plugin Flowy indexes from it ships 27 components: 26 skills and 1 MCP server. Flowy indexes what the repository publishes; it does not own, host, or run any of it.

The everyday memory commands

This group covers the basic verbs. remember stores a memory verbatim when you say something like remember this, save this, or note that, classifying it with a type and metadata as it goes. search and peek handle retrieval, and their descriptions are close to identical: both search memories and return compact one-liner results, or look up a single memory by ID, useful for a quick check on whether something was recorded or for resolving a citation. forget deletes a memory by search query or by ID, asks for confirmation first, and can undo a recent addition. pin protects a specific memory, an architecture decision or a security constraint for example, from later removal during cleanup. A related skill, memory-triage, runs a similar evaluate-and-store pass, though its own description ties it to a separate project called openclaw-mem0, closer to a cousin of remember than a distinct command.

Keeping stored memory clean

Memory quality gets its own set of skills. memory-reviewer audits what is already stored, flagging duplicates, contradictions, and stale entries with recommendations, meant to run before any cleanup. dream is that cleanup step: it merges duplicates, resolves contradictions, and prunes stale entries once the memory count is high or search results start feeling noisy. memory-dream covers similar ground but goes further: it also strips noise and credentials, rewrites unclear entries, enforces expiration on old memories, and can apparently trigger automatically, not just on request. stats reports usage: counts by category, an age breakdown, and API latency. tour browses everything already stored, grouped by category with full detail, good for an overview rather than a search.

Working across more than one project

context-loader runs at the start of a task, searching for relevant memories and injecting them before you begin, surfacing past decisions and coding conventions unasked. list-projects lists every project with stored memories, showing counts and last activity, and switch-project points reads and writes at a different project, or turns on search across every project and user at once. policy is where the extraction rules live: what mem0 should remember or ignore for a given project, kept in a project file called mem0.md. export writes everything to a portable markdown file for backup or handoff, and import brings memories back from that file, or from Claude Code's MEMORY.md, for a new project or a restore.

Setting up and checking the connection

onboard handles first run setup: API key configuration, MCP authentication, pulling in existing project files, and setting up coding categories. health and status both diagnose the same problem, connectivity, API key validity, and whether reads and writes work, and their descriptions barely differ, so either works when a search comes back empty or a save fails. The plugin's one MCP server is also named mem0. Flowy's index carries no separate description for it, but the skills above check MCP connectivity and authentication by name, so it looks like the direct path to the same memory operations the others wrap as commands.

Wiring mem0 into your own application

The remaining skills are not about managing memory inside a session at all; they are reference material for a developer adding Mem0 to their own product. mem0 documents the Mem0 Platform SDK, firing on mentions of a MemoryClient or a persistent memory layer. mem0-cli covers the separate command line interface, for typing mem0 commands such as mem0 add or mem0 search directly in a terminal. mem0-vercel-ai-sdk documents the Mem0 provider for the Vercel AI SDK. mem0-integrate runs a guided, test-first pipeline that adds Mem0 to an existing repository, detecting its language and asking you to choose between the hosted Platform and the self-hosted option. mem0-test-integration is the companion check, run afterward in the same workspace: it installs dependencies, runs the repository's own tests, exercises a real end-to-end flow against your API key, and produces a scorecard. mem0-oss-to-platform plans and carries out a migration from the self-hosted SDK to the hosted one.

How do I install it

Install from the listing page linked at the top of this guide. Flowy indexes mem0ai/mem0 from its public repository under the data category. It does not own, host, or modify the project, and all 27 components above stay on the author's own releases in the source repository.