AI memory is getting f...cking illegal 10 open-source GitHub projects that stop agents from starting from zero EVERY new session 01 Mem0 ▸ → 66K+ stars 02 Hindsight ▸ → retain → recall → reflect 03 memU ▸ TURN MEMORY INTO KNOWLEDGE 04 Cognee ▸ → documents + code... + conversations → knowledge graph 05 Graphiti ▸ → remembers how facts change over time 06 OpenViking ▸ MAKE THE AGENT STATEFUL 07 Letta ▸ → memory + identity across sessions 08 Letta Code ▸ REMEMBER ACROSS THE STACK 09 OpenMemory ▸ 10 Agent Memory Benchmark ▸ the loop: experience → remember → connect → retrieve → act → update 3 builds I'd actually test: coding: Hindsight → Cognee → Letta Code personal agent: Mem0 → Graphiti → Letta company brain: Cognee → Graphiti → Hindsight and this is where the sh...t gets interesting: bigger context isn't the same as memory a model can hold 1M tokens and still meet you for the first time every morning these projects change that save this before your agent forgets it ⭣show more

Clean lines and clear structure in this terminal

saving user context between sessions always made my ai agents way less annoying to work with

this is what I was looking for

66k stars on mem0 is wild

the tenth github repo is the most versatile

I'll keep this selection for myself.

This is one of the biggest unlocks for agentic AI. 🧠⚡️ An agent that can remember, learn from previous sessions, and build context over time feels less like a stateless tool and more like a persistent digital collaborator. The memory layer may become just as important as the model itself. 👀

saved that github library

wow, pure gold how did find it?

A bigger context window still doesn’t give an agent a memory.

These repos are worth reading

What is this

Mines not released yet

The context vs memory distinction is the real point here. Most teams just keep raising token limits instead of building retrieval that persists.

A million-token window still cannot replace retrieval that updates durable facts.

which one actually remembers without me building a whole pipeline

Very well put

There’s also it’s #1 on BEAM and can be self-hosted

so usefull thx bro

Love this take

Honcho needs to go on that list imho.

@grok whats the performance of daidocs against all the 10 he mentioned?

The interesting part isn’t just persistence, it’s deciding what an agent should forget. Memory without deletion controls gets messy fast.