Claude Remember plugin

A plugin that saves Claude Code sessions, compresses them through Haiku into daily summaries and loads them at session start

  • 211 ★ on the repository
  • Last commit
Plugin

Extended access

Why this level

  • Hooks run with your shell privileges on tool calls, and the nested `claude -p` inherits your environment and login
  • Session summaries are written to files and may contain sensitive project data, and with a connected git remote they are pushed to its server
All reasons and checks

digital-process-tools/claude-remember

Install

Manual install

/plugin marketplace add Digital-Process-Tools/claude-marketplace

Then `/plugin install remember@dpt-plugins` and restart Claude Code: hooks are read at session start.

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • Hooks run with your shell privileges on tool calls, and the nested `claude -p` inherits your environment and login
  • Session summaries are written to files and may contain sensitive project data, and with a connected git remote they are pushed to its server

This is third-party code. Review the repository files before installing.

What it does

Claude Remember hooks into the coding agent's lifecycle and keeps layered memory. While you work, `save-session.sh` extracts the session, Haiku condenses it into short lines, hourly compression turns those into a day summary, and daily consolidation moves summaries into `recent.md` and `archive.md`. Each layer is shorter than the one above, so the needed context fits in a small number of tokens. At session start the `SessionStart` hook injects memory files into the agent's context by itself: `identity.md` (who the agent is), `remember.md` (a handoff note from the last session), `now.md`, today's summary, the last seven days and the archive. After a context compaction only `identity.md` is injected again. Two commands are manual. `/remember` before clearing context or ending a session writes a short note for the next session. `/remember:doctor` prints paths, storage mode and whether each hook has fired for this project. It works in Claude Code and Codex, and Antigravity CLI has a separate installer.

Who it is for. Developers who return to the same projects in Claude Code every day and are tired of re-explaining yesterday to the agent.

Good fit when

  • Every session starts from zero and the agent repeats mistakes already discussed yesterday
  • Before `/clear` or at the end of the day you want a note for the next session about where you stopped
  • Several git worktrees serve one project and memory must be shared by all of them
  • You want decision history as plain markdown files you can read, search and delete
  • Memory is not appearing and you need to find out whether hooks fired in the project

Not a fit when

  • There is no Claude CLI access with Haiku: compression runs through it
  • The project is commercial and you want to bundle the plugin into a product you sell: the license forbids commercial redistribution
  • Sessions contain secrets or personal data that must not be stored in summaries
  • You need semantic memory with meaning-based search and a database: this keeps flat files compressed by time

Example request

/remember

Limitations

It needs Python 3.9, the Claude CLI with Haiku access, Bash 3.2, `jq` and coreutils. On Windows it requires a POSIX shell (Git Bash, MSYS2 or WSL), and most of the test suite is skipped there. One saved session costs under a cent and a day of work comes to a few cents.

The plugin runs with your shell privileges like any hook: the summarizer launches a nested `claude -p` that inherits your environment and login. Memory is stored locally in `.remember/` inside the project. Git backup triggers when the external store sits inside a git repository with a remote, and then summaries are pushed to that server. The Community license forbids commercial redistribution and competing use.

How to disable. Disable or uninstall the plugin in Claude Code with `/plugin` and restart the session. The data stays in the project's `.remember/` folder (or in `~/.remember/<slug>/`); delete it by hand if you no longer need the memory. The promo line about sibling plugins is switched off with `"features": {"plugin_promos": false}` in `config.json`.

FAQ

How do I install Claude Remember?

In Claude Code run `/plugin marketplace add Digital-Process-Tools/claude-marketplace`, then `/plugin install remember@dpt-plugins` and restart. For Codex use the pair `codex plugin marketplace add Digital-Process-Tools/claude-remember` and `codex plugin install remember`.

Where is the memory stored?

By default in the `.remember/` folder inside the project, in external mode in `~/.remember/<slug>/`. These are plain markdown files: `now.md`, daily summaries, `recent.md` and `archive.md`. Memory is tied to the main checkout, so all git worktrees share one.

How much does it cost?

Compression runs through Haiku, the cheapest Claude model. Saving one session costs under a cent and a full day of work a few cents. The same Claude CLI login pays for the plugin's calls.

What if memory does not load?

Run `/remember:doctor`: it prints resolved paths, storage mode and whether each hook has fired for this project. After installing the plugin, restart Claude Code.

What does the plugin send out?

Nothing in the default install: memory stays local. The exception is git backup, which turns on by itself when the external store's parent folder is a git repository with a remote. Summaries then go to that server.

Discussion

No questions or reviews yet.

Plugins
Official

The Supermemory plugin for Claude Code: persistent memory across sessions with automatic prompt recall, conversation capture and a codebase index

PluginBroad access2.8KRepository stars

A Claude Code plugin that gives the agent a name, a personality and memory across sessions, plus chat channels.

PluginBroad access64Repository stars

hipocampus gives an agent file-based memory: a 3,000-token topic index loads each session and shows what came up before

PluginExtended access211Repository stars

Claude Code and Codex hooks search your notes and surface a known answer before a reply and after a failed command.

CLIExtended access88Repository stars

A long-term memory plugin that keeps decisions, preferences and task state across sessions and injects only what a query needs

PluginBroad access401Repository stars

A persistent memory plugin: stores observations from agent sessions and injects relevant context into new ones

PluginBroad access116.5KRepository stars

Comparison with similar tools

Cards that share the main search query, and how they differ.

ToolOfficialAuthenticationRepository starsMain difference
Claude Remember (this card)no211
Claude-Memno116,457a persistent memory plugin that stores observations in a database and injects relevant context
agentmemoryyesnot required30,314local memory with hooks and access to past decisions through MCP

README in short

A hook table maps Claude Code, Codex and Antigravity events to the plugin's scripts: memory loading, a timestamp in the prompt, saving after a batch of tool calls and a write at session end. Details live in separate documents: configuration, diagnostics, handoff between sessions, Windows notes, external storage and git backup security.

README badge

Report a problem