Claude Code hook

UnseveredMemory

Simple, organic, lightweight yet purposeful memory system for Claude Code.

49 stars 3 forksHooks Plugin marketplace Source

Install

Inside Claude Code
/plugin marketplace add blas0/UnseveredMemory

Found in the project README. Check it for setup steps after installing.

From source
git clone https://github.com/blas0/UnseveredMemory.git

From the README

Unsevered Memory

A markdown-based memory system for Claude Code with enforced persistence.

Changelog

- Added support for Claude plugins + npx packaging

What Makes This Different

Most memory systems inject context at session start and hope Claude remembers. This doesn't work because:

1. Context compaction loses early instructions
2. Claude has no obligation to follow suggestions
3. Long sessions forget the protocol

Unsevered Memory solves this with enforcement:

The UserPromptSubmit hook survives context compaction by being injected fresh on every message.

Architecture

Installation

Choose your preferred method:

Option A: Claude Plugin (Recommended)

Then per-project:

Option B: npx

Option C: Manual

Then per-project:

What Gets Installed

Global (~/.claude/):
- CLAUDE.md - Global memory protocol
- settings.json - Hook configuration (3 hooks)
- hooks/ - memory-load, memory-remind, memory-save
- skills/orchestrate/ - Workflow instructions
- commands/orchestrate.md - Orchestrator command

Per-Project:

Workflow

Session Start
1. Hook loads context.md and scratchpad.md
2. Claude sees current state and any unfinished work
3. Hook hints about .ai/ documentation

During Session
Every prompt shows:

Claude writes to scratchpad.md as it works:

Claude updates .ai/ when patterns emerge (3+ uses).

Session End
1. Hook archives scratchpad to sessions/YYYY-MM-DD.md
2. Hook reminds to update context.md
3. Claude updates context with current state

Orchestrator Mode

For complex multi-step tasks:

The orchestrator:
1. Reads all memory files
2. Breaks task into subtasks
3. Delegates to specialized agents
4. Updates memory after each step
5. Never loses context

File Purposes

Repository Structure

Philosophy

- Enforced - UserPromptSubmit hook survives context compaction
-…
Read the full README on GitHub

How hooks work

  • Hooks are shell commands Claude Code runs at fixed points: before or after a tool call, when you submit a prompt, when Claude stops, on notifications and at session start.
  • They are set in the hooks section of ~/.claude/settings.json (or .claude/settings.json for a project), or come bundled in a plugin.
  • A hook can block a tool call, add context, or run a formatter or test. Read the scripts before you install them: they run on your machine with your permissions.
Claude Code docs

Details

Type
Hooks
Language
Shell
License
MIT
Last updated
3 January 2026
Topics
ai-codingclaudeclaude-codedeveloper-toolshooksmarkdownmemorypersistenceproductivity

FAQ

What is UnseveredMemory?

Simple, organic, lightweight yet purposeful memory system for Claude Code. It is a hook for Claude Code, from blas0/UnseveredMemory on GitHub.

How do I install UnseveredMemory?

Run /plugin marketplace add blas0/UnseveredMemory inside Claude Code, then follow the setup in the README.

Is UnseveredMemory popular?

blas0/UnseveredMemory has 49 GitHub stars and 3 forks. It was last updated on 3 January 2026.

More hooks

Browse the most popular Claude Code mods, all hooks, or the skills and MCP server directories.