knowledge graphs are INSANELY good at giving agents human-like memory! today, we're building an MCP-powered memory layer that can be shared across all your AI apps like Cursor, Claude Desktop etc. it's built using a real-time knowledge graph. 100% open-source and self-hosted.

tech stack: - @zep_ai 's Graphiti as a memory layer - @neo4j to store the knowledge graph - @Docker to self host the MCP server find all the code here:

@neo4j @Docker If you found it insightful, reshare with your network. Find me → @akshay_pachaar ✔️ For more insights and tutorials on LLMs, AI Agents, and Machine Learning!

solving a really good problem here using knowledge graphs and memory. I think most people use AI apps in silos and there's no simple way to automatically connect them via a shared memory.

Absolutely, everyone can benefit from this. Shared memory has a tons of use cases.

Knowledge graphs for AI memory; brilliant. Maybe my to-do list will finally remember itself.

What about latency? When your graph has multiple nodes and relationships, it takes more time to retrieve high quality result and rerank it. How do you deal with that? That’s my biggest issue with graphs atm

This is a huge step forward—shared AI memory is game-changing.

What other use cases do you have in mind ?

The open-source aspect is a big plus, but I wonder about the scalability of a self-hosted solution for many users.

Scalability is not an issue, you can always deploy this container in a any scalable cluster.

real-time memory sharing? creators could finally breathe!

Could you suggest best RAG solution that supports AWS Bedrock ?? Thanks in advance!!!

This looks amazing! Can’t wait to see AI apps with shared memory.

dude, thanks

@AlphaLebowski You're welcome! Cheers! :)

A shared memory layer changes the game. Agents that can retain and connect experiences across apps start to feel less like tools and more like collaborators. It opens up the question of what new kinds of workflows become possible when memory is no longer siloed.

Are we just rebuilding search engines?

this is a solid approach to breaking down silos. shared memory really can change how we use ai apps together. any plans for more integrations down the line?

loved it bro. thanks

Glad you liked it! Cheers! :)

should try it with helix, we have a zep integration

@akshay_pachaar When would you use Graphiti vs MindsDB for real-time distributed data like social media posts and interactions or crypto / blockchain transactions and trends? BTW, your content is amazing, thank you!

Knowledge graphs are great for this! LATRACE takes it further—event-centric where time is the anchor, state changes are first-class. Text/video/audio in same pipeline → knowledge graph. SOTA on temporal reasoning.

Knowledge graphs + semantic memory is the move. We built similar at Remembra — bi-temporal versioning so you can query what the agent knew at any point in time. Critical for debugging multi-agent workflows. Also MCP-native, self-hosted. Different architecture but same goal: agents that actually remember.

Knowledge graphs drastically reduce AI agent hallucinations. True reliability demands structured context and explicit, traceable verification checks at every workflow step, not just final output.

knowledge graphs for AI are what spacetime is for consciousness, a living network of stored relationships, updating itself in real time. #MST #MultiSimulationThesis

Come show us live on @promptorgtfo? CC: @gadievron