Claude Code memory that belongs to you

As of 2026-11-21, PLUR 0.23.1 (@plur-ai/cli and @plur-ai/mcp, both requiring Node.js 22 or newer).

Claude Code already remembers some things. Its CLAUDE.md files hold the instructions you write, and its auto memory saves notes Claude decides are worth keeping. PLUR adds a memory layer that sits outside Claude Code: corrections, preferences and project conventions are saved as plain YAML in ~/.plur/ on your disk, found again by local search, and injected into new sessions by hooks. You can read every memory, correct it, or retire it with one command. The same store works in Cursor, Codex and other agents, so what you teach Claude Code is there when you switch tools.

Install

You need Node.js 22 or newer.

Prefer to let your agent do it? Tell it: "Go to plur.ai and install memory." It follows plur.ai/install.txt: checks for Node.js 22, runs the npm install and plur init, verifies with plur doctor, and tells you what only you can do (restart).

1. Install the CLI and the MCP server, then set up Claude Code:

npm install -g @plur-ai/cli@latest @plur-ai/mcp@latest
plur init
plur doctor

plur init creates the store at ~/.plur/, registers PLUR as a user-scoped MCP server in ~/.claude.json, and writes global hooks to ~/.claude/settings.json. The hooks inject memory when a session starts and close session state when Claude Code ends a session. plur doctor checks the setup and the MCP handshake, and downloads the local search model if it is missing. Without the model, PLUR falls back to keyword search.

2. Restart Claude Code.

3. Say yes to memory in this folder. The first time PLUR runs in a folder you have not decided about, it asks once whether to use memory there. Answer in the chat, or run:

plur folders set "$PWD" --on

4. Check it works. Ask Claude to call plur_status. It should report an engram count and the storage path.

Run plur init again after each upgrade so the hooks and MCP entry match the new version.

Several projects? Give each its own scope, so project memories stay apart while your global preferences follow you everywhere:

cd ~/projects/my-api
plur init --domain myapi --scope project:my-api

What it remembers, and how to read, correct and delete a memory

PLUR stores engrams: short statements such as a correction ("never force-push to main"), a preference, or a convention, each with a scope, a type and a strength that grows with use and fades when unused. It is agent memory, not a code index: it does not parse or search your source files.

A worked example. In a session you tell Claude:

"Use pnpm in this repo. npm install breaks the lockfile in CI."

Claude saves it with plur_learn, and it lands in ~/.plur/engrams.yaml with an id such as ENG-2026-11-17-001. Next week, in a new session, you ask how to install dependencies. The hooks inject the memory and Claude answers with pnpm without being reminded.

Read it. Every memory is plain text, so pick whichever view you like:

plur list                                  # every engram
plur recall "how do I install dependencies" # search, the way the agent does
plur dashboard                             # local, read-only page in your browser

Or open ~/.plur/engrams.yaml in any editor.

Correct it. The team moved to bun. Retire the old memory and teach the new one:

plur forget ENG-2026-11-17-001 --reason "moved to bun"
plur learn "Use bun in this repo, not pnpm or npm"

If a memory keeps showing up where it does not help, rate it down instead: plur feedback ENG-2026-11-17-001 negative.

Delete it. plur forget <id> retires the memory: it is no longer recalled or injected. The entry stays in engrams.yaml, marked status: retired, so you can still see what was retired and when. Because the store is a plain file, you can also remove an entry by hand.

To turn memory off for one folder entirely: plur folders set ~/private-project --off.

How it compares with Claude Code's built-in memory

Claude Code's own memory is good, and readable too: auto memory is plain markdown you can open with /memory, edit or delete. Keep using CLAUDE.md for the rules that must always apply. PLUR does not replace either. It adds a store that is not tied to one tool or one repository.

CLAUDE.md Auto memory PLUR
Who writes it You Claude The agent, through PLUR's tools, or you from the CLI
Where it lives Project, user or managed files ~/.claude/projects/<project>/memory/, one per repository ~/.plur/engrams.yaml, one store for all projects, separated by scope
What loads each session The whole file First 200 lines or 25KB of MEMORY.md; topic files on demand Memories selected by relevance and injected by hooks
Search None needed: it is always loaded Claude reads topic files when it needs them Local hybrid search (keyword plus embeddings), no API calls
Read and edit Any editor, /memory Plain markdown, /memory Plain YAML, plur list, plur dashboard
Remove Edit the file Edit or delete the files plur forget <id>, or edit the file
Learns what helps No No Feedback ratings and decay of unused memories
Other machines Through git, if committed Machine-local plur sync to a private git remote you choose
Other tools Claude Code (other tools read AGENTS.md) Claude Code only Cursor, Codex, OpenCode, OpenClaw, Hermes, Antigravity and any MCP client

FAQ

Doesn't Claude Code already have memory? Yes. CLAUDE.md holds instructions you write, and auto memory holds notes Claude saves for itself, one directory per repository on one machine. PLUR is for memory that should outlive one tool, one repository or one machine, and that you want to search, rate and retire memory by memory.

Should I keep my CLAUDE.md? Yes. Put rules that must apply in every session in CLAUDE.md. plur init adds a short PLUR section there so Claude knows when to save and look up memories. PLUR holds the corrections and conventions that build up while you work.

Do my memories leave my machine? Not by default. The store is local YAML and search runs locally; plur doctor downloads the search model once. Syncing is opt-in: plur sync pushes to a git remote you name, which should be private.

Will Cursor and Codex see what I taught Claude Code? Yes, once PLUR is set up in them (plur init --cursor, plur init --codex). They read and write the same ~/.plur/ store.

How do I stop PLUR in one project? Run plur folders set <folder> --off. The hooks and the memory tools then do nothing in that folder.

What if I want to leave PLUR? Your memories are a YAML file in an open, published format. Copy it, read it, or delete ~/.plur/. For what to test before trusting any memory tool with your data, see Test Your Exit Plan First.

Related: How to Add Persistent Memory to Claude Code · How to Recover When an AI Agent Remembers the Wrong Thing · Stop Agent Memory Loss: A Cross-Session Verification Checklist

Get started

npm install -g @plur-ai/cli@latest @plur-ai/mcp@latest && plur init

Open source, Apache-2.0: github.com/plur-ai/plur