Claude-Mem
AI MEMORY THAT ACTUALLY WORKS

Stop explaining context.
Start building faster.

Claude-Mem is your AI's trusty note-taking sidekick. Never lose track ever again.

Quick Install:
/plugin marketplace add thedotmack/claude-mem&&/plugin install claude-mem
Scroll to discover

One AI takes notes about what another AI does.

Claude-Mem watches your AI coding assistant work and captures what matters. Every decision, every bug fix, every architectural choice — remembered automatically.

BeforeWhat happened in previous sessions
CurrentPresent context and recent work
NextWhat comes next, pending tasks
Live Observation

Your AI doesn't have to remember anymore

A dedicated observer AI watches every session, generating searchable observations in real-time. Complete logs with before-and-after context, organized by time—just like how humans handle the same problem.

🔍Searchable by time
📸Before & after context
Generated live
claude—mem

How you'd remember it—
for your AI

Every observation is auto-categorized. Filter by decisions, bugfixes, features, or discoveries. Combine with file scope for surgical precision.

type:decision file:auth.ts
⚖️decision47
🔴bugfix23
🟣feature89
🔵discovery156
By filedecisions for src/auth/index.ts
or
By conceptdecisions about "token refresh"

File & Concept
Scoping

Query by file path or semantic concept. “What decisions affected index.ts?” or “What do we know about auth?” Both work.

Progressive
Disclosure

Sessions start with a lightweight index—titles, types, timestamps. The LLM fetches full observations only when it needs depth.

Token-efficient by default, never shallow when it matters.

Session start
⚖️JWT over sessions for auth~40 tokens
🔴Race condition in token refresh~35 tokens
...48 more observations~2.1k total
On demand
Full observation~850 tokens
7 before
🔵 Researched auth approaches🔵 Compared session vs token...
match⚖️ Chose JWT for statelessness
7 after
🟣 Implemented token service🔴 Fixed refresh race condition...

Before/After
Context

Every observation includes what came before and what followed. The LLM sees causality—not snapshots.

“Why did this decision lead to that bug?” becomes answerable.

Coming Soon

Real-Time Agent Data

An open standard for AI agent memory.
RAG captures knowledge. RAD captures intelligence.

Just as RAG (Retrieval Augmented Generation) standardized external knowledge retrieval, RAD will standardize how agents capture and retrieve their own working memory. Hook-based architecture, intelligent compression, temporal awareness — all in one open protocol.

Quick Install:
/plugin marketplace add thedotmack/claude-mem && /plugin install claude-mem