Claude Code + MemorySync

Overview

Install the MemorySync plugin from its marketplace and Claude Code remembers: lifecycle hooks send every prompt to fact extraction the moment you send it (only the durable facts in it are stored, and replies are not sent) and inject relevant memories at session start, alongside each prompt of 24 characters or more, and again after context compaction. Facts are tagged with the git repository (worktree-aware) while your preferences follow you everywhere, the full MemorySync MCP server and memory skills come bundled, and every hook exits 0 on every failure — memory being down can never break a session. Cross-platform Node hooks: Windows, macOS and Linux natively. The same plugin works in Claude Cowork cloud sessions via repo-declared settings.

Setup status and requirements

Supported version

memorysync@memorysync plugin 1.4.0

Last setup review

2026-10-01

Permissions

A MemorySync API key with read and write scopes (or OAuth through the bundled MCP server).

Limits

Requires Claude Code 2.0+ and Node.js 18+ on PATH for the hook scripts. Claude Cowork requires the repo-declared settings file and a Cloud Environment API key. Claude Desktop’s local Cowork tab does not run plugin hooks — use the MCP server there.

Capabilities

• Automatic capture: every prompt sent for fact extraction by lifecycle hooks, not model discretion
• Automatic recall at session start, per prompt (24+ characters), and re-injected after compaction
• Per-repo project scoping from the git remote, worktree-aware
• Cross-platform Node hooks — native Windows, macOS, Linux
• Session-safe by contract: every hook exits 0 on every failure
• Bundled MemorySync MCP server (API key or OAuth) + docs server
• /memorysync:status, /memorysync:remember, /memorysync:recall

Quick Start

# In your shell (one time):
export MEMORYSYNC_API_KEY=ms_...

# In Claude Code:
/plugin marketplace add memorysyncio/memorysync-plugins
/plugin install memorysync@memorysync

# Verify:
/memorysync:status

Use Cases

Explore More