Back to browse
I made an open-source memory layer for agents

I made an open-source memory layer for agents

by MarsB·May 22, 2026·8 points·0 comments

AI Analysis

●●SolidSolve My ProblemShip It

Graph memory with auto-extraction beats simple vector stores for context.

Strengths
  • Neo4j-backed graph storage enables relationship traversal, not just keyword matching.
  • Built-in Qwen2.5 pipeline auto-extracts entities and links memories without manual tagging.
  • Dual deployment: self-hosted Docker for control or managed cloud for quick setup.
Weaknesses
  • Competes directly with established players like Mem0, LangMem, and Zep in a crowded space.
  • Reliance on a specific 7B model for extraction may limit customization for advanced users.
Category
Target Audience

Developers building stateful AI agents or LLM applications

Similar To

Mem0 · Zep · LangMem

Post Description

Store memories, auto-extract entities and relationships, search semantically. MCP server + REST API + SDKs. Self-hostable, cloud option, MIT license.

Similar Projects

AI/MLMid

Magma Memory Claude Browser cloaking tool

Local AI memory with REM compression, but gated behind a contact form.

Bold BetSlick
vektormemory
202mo ago
Developer Tools●●Solid

Agents Remember – Git-aware memory for coding agents

Git-verified memory ledger beats vague agent context files.

Big BrainNiche Gem
foxfire_1st
3213d ago