seanGSISG

Always On Memory Agent — AI skill for Claude Code

AI community

Always-on memory for Claude Code.

How to install Always On Memory Agent

This entry records only its repository, not the path inside it, so there is no exact command to give. Open seanGSISG/always-on-memory-agent and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Always On Memory Agent does

Always-on memory for [Claude Code](https://docs.anthropic.com/en/docs/claude-code). Automatically records sessions, extracts structured observations via LLM, consolidates cross-session insights, and i

Alternatives in AI

  • Codex Plugin — 's /codex:adversarial-review command 9.6k ★
  • Tma1 — Local-first observability your agent reads back 116 ★
  • Cross Agent Session Resumer — Resume AI coding sessions across providers: converts Codex, Claude, Gemini, and other session formats through 71 ★

README

claude-memory

Always-on memory for [Claude Code](https://docs.anthropic.com/en/docs/claude-code). Automatically records sessions, extracts structured observations via LLM, consolidates cross-session insights, and injects relevant context when you start a new session — so Claude remembers what you've been working on.

What It Does

Every Claude Code session generates valuable context — architectural decisions, bug fixes, failed approaches, dependency quirks. Without memory, each session starts from scratch. `claude-memory` fixes that.

The Memory Loop

Session Start ──► Inject relevant memories into context
       │
  Work happens (tool calls, edits, searches recorded)
       │
  Session End ──► Extract observations via LLM (background)
       │
  Consolidate ──► Find cross-session patterns, resolve contradictions (background)
       │
  Next Session ──► Inject again, now with richer context

What Gets Remembered

Observations are extracted with priority levels:

Priority Category Examples
P1 Critical Architectural decisions with rationale, security-sensitive patterns
P2 Important Bug fixes with root causes, dependency quirks and workarounds
P3 Useful Code patterns/conventions, file relationships, test strategies
P4 Minor Failed approaches, dead ends, environment setup details

Each observation includes entities (files, functions, packages), topic tags, and an importance score that decays over time — so stale memories fade naturally while critical decisions persist.

Architecture

claude-memory-plugin/src/
├── .claude-plugin/
│   └── plugin.json              # Plugin manifest (v0.2.0)
├── hooks/
│   ├── hooks.json               # Hook event registrations
│   ├── lib.sh                   # Shared shell functions (venv resolution)
│   ├── session-start.sh         # Register session + inject memories
│   ├── session-end.sh           # Finalize session