Front page / Concepts / Memory
Memory
By default every chat starts from nothing. Memory lets the assistant carry forward what it learned about you and your work: your role, your product, your preferences, the decisions already made. It can be the app's built-in memory, a file you maintain, or a dedicated memory server; in all cases you decide what it remembers.
Where it stands
Among the 14 concepts. Bars: GitHub stars gained this week.
- MCP and tool use+3.9k
- Agentic loops+1.3k
- Knowledge graphs (GraphRAG)+1.2k
- Retrieval (RAG)+1.2k
- Prompt engineering+1.1k
- Vibe coding+980
- Workflows, not agents+565
- Spec-driven work+16k
- Memory+7.3k
- Agent skills+5.7k
- Multi-agent orchestration+5.6k
- Computer use+4.7k
- Context engineering+3.3k
- Evals and verification+929
How to use it today
In ChatGPT or Claude, as a person doing the work, not as a developer.
- 1Turn on memory in ChatGPT or Claude and tell it, once, who you are and what you work on; then check the memory list and delete what is wrong.
- 2In Claude Code, keep a CLAUDE.md with the standing facts of the project: what it is, the rules, the decisions made and their dates.
- 3Write decisions into memory explicitly ("remember that we chose quarterly pricing on 2026-09-30") rather than hoping it is inferred.
- 4Review the memory monthly; outdated memories cause the same errors as outdated documents.
- 5Keep personal and client information out of memory unless you know where it is stored.
What changes day to day
You stop re-introducing yourself and your project. Answers start from where you left off, and the assistant stops suggesting what you already rejected.
Where it goes wrong
Memory that is never pruned fills with stale preferences and one-off instructions that the assistant then applies forever. Shared or team memory needs an owner.
Listings that use it well
Chosen by hand after reading them, each with the reason. Install from the listing's own page.
| Listing | Kind | Why it is here | Stars |
|---|---|---|---|
| claude-mem | Agent skills | Captures every session, compresses it and injects relevant context into future sessions. | 115k |
| ai-agent-book | Agent skills | memory-system covers memory tiers, storage formats, Mem0 versus Memobase and when memory beats RAG. | 57k |
| agentmemory | Agent skills | Persistent memory for coding agents with hooks, recall, remember and session history. | 30k |
| Productivity (Knowledge Work Plugins) | Agent skills | memory-management keeps a two tier context library of people, projects and glossary so Claude stops asking. | 35k |
| mempalace | Agent skills | A local memory system with semantic search and a temporal knowledge graph. | 60k |
| cognee | Agent skills | Memory platform that turns data into graph plus vector memory for agents. | 33k |
Elsewhere
Repositories that embody the idea and are not in the catalog.
- mem0ai/mem0 ↗Drop-in memory layer for agents and apps.
- letta-ai/letta ↗Stateful agents with self-editing memory (from MemGPT).
- getzep/graphiti ↗Temporal knowledge graph engine for agent memory.
Standing
Our judgment, with the reason and the sources it rests on.
Rising since 2025-04. Both ChatGPT and Claude shipped persistent memory in 2025 and memory files are now standard in the coding assistants; what to remember is the open design question. ChatGPT began referencing all past chats in April 2025, Claude got memory for Team and Enterprise in September 2025 and for Pro and Max in October 2025, and Claude Code turned on auto memory files by default in February 2026. Claude made memory free for all users in March 2026, with a tool to import memories from other assistants. Adopted faster each month; the vendors are building for it and new tools keep appearing.
- OpenAI on X: memory in ChatGPT can now reference all of your past chats2025-04-10
- Bringing memory to Claude2025-09-11
- Release v2.1.59 (Claude Code: auto memory)2026-02-26
Related: Context engineering, Retrieval (RAG), Knowledge graphs (GraphRAG)