• 

  MemgenX - Universal Memory Layer

MemgenX: Universal Persistent AI Memory Layer

Inspiration

AI is becoming part of every workflow, yet every LLM and agent forgets everything once a session ends. Users repeat the same context across ChatGPT, Claude, Gemini, Cursor, Kiro, and countless tools wasting time, tokens, and productivity. We wanted to fix the biggest gap in AI today: the lack of persistent, shared memory across all models and apps.

What it Does

MemgenX provides a universal, persistent memory layer for all LLMs, AI agents, and AI applications. It automatically stores user facts, preferences, history, and relationships, then recalls them whenever any model or agent needs context. This eliminates repeated prompting, reduces token costs, and ensures consistent, personalized AI everywhere.

How We Built It

• Chrome Extension to capture prompts & responses across AI tools.
• Centralized backend with memory storage, embeddings, and retrieval.
• RAG-based memory engine that stores structured facts & semantic vectors.
• Multi-model compatibility (any LLM, any agent, any IDE).
• Dashboard web app with analytics: memory usage, LLM usage, API logs.
• MCP-compatible design so any MCP-based AI app (Cursor, Claude Desktop, Kiro, Windsurf, etc.) can use MemgenX as a tool server.
• AWS infrastructure for scaling: serverless backend + vector storage.

Challenges We Ran Into

• Designing a consistent memory representation that works across different LLMs.
• Deciding what to store, what to summarize, and how to avoid “memory pollution.”
• Ensuring memory retrieval is fast enough to support real-time AI interactions.
• Making the Chrome Extension reliably capture content across multiple AI sites.
• Building a system that’s generic enough to work for any user, any model, any platform.
• Balancing privacy, consent, and user control for stored memory.

Accomplishments That We're Proud Of

• Built a working universal memory layer MVP in a very short time.
• Created a Chrome extension that works with multiple LLM platforms from day one.
• Designed a scalable architecture ready for real-world AI workloads.
• Achieved end-to-end memory flow: capture → embed → store → retrieve → inject.
• Positioned to form MemgenX as one of the first MCP-ready memory tools for AI ecosystems.
• Built a clean web dashboard for analytics and memory management.

What We Learned

• Persistent memory dramatically improves AI usability and consistency.
• Cross-LLM workflows are the future, and memory must be model-agnostic.
• Users want full visibility and control over what AI remembers.
• MCP (Model Context Protocol) is the emerging standard for connecting tools to LLMs.
• Building a universal memory layer requires thoughtful design, not just vector search.
• Memory is not just storage, it’s a product experience.

What's Next for MemgenX

• Full MCP Tool Server release for Cursor, Claude Desktop, Windsurf, Kiro, Replit, etc.
• Mobile app for on-the-go memory capture (voice → memory).
• VSCode extension + desktop app for developer workflows.
• Team & enterprise memory with shared namespaces.
• Self-hosted / VPC deployments for businesses.
• Fine-grained memory approval pipeline (“AI inbox” for user-curated facts).
• Faster embeddings pipeline + semantic pruning for clean, high-quality memory.

Note : I have provided the chrome extension zip file unzip it

        -  Upload folder to the extension tab

        - Login with google account to use it

Built With

• fastapi
• nextjs
• python
• render
• s3
• stripe
• supabase
• typescript
• vercel