Cortex Protocol V1.5 Update + V2 Architecture + Launched on Product Hunt... would love your upvotes!

Hey Gemini 3 team, judges & fellow hackers,

Since my original 200-word submission (word limit cut a lot of detail), I’ve shipped Cortex V1.5 with real post-hackathon upgrades:

* Brand new landing page with full documentation, step-by-step setup guide, improved UI(used Gemini 3.1 Pro in Antigravity to redesign the UI):
https://cortexmcp.vercel.app

* Memory Transfer feature (the one I really wished existed during the hackathon):
- Copy the provided prompt from Cortex
- Paste it into ChatGPT / Claude / Gemini / Cursor / any AI
- The AI returns everything it knows about you
- Paste that raw output back into Cortex
- Cortex instantly turns it into structured, queryable memories across Hot / Warm / Cold layers that any AI agent can use forever.
No more repeating context!

* Improved MCP server (faster hybrid search, better GraphRAG, Sovereign State encryption)

* Full PAT system: you still fully control what gets stored, shared, or hidden.

I’ve also already designed the new V2 architecture:
- Costs <$1/month per user at moderate usage
- Significantly improved ingest & retrieval speed
- Can store years of memories without any quality degradation

On the competition side... SuperMemory, Zep, and OpenMemory are all excellent projects and I respect the work they’re doing. But most of them are built for developers (app-first or agent-first memory layers where the control stays with the app or the builder).

Cortex Protocol flips that completely:
It’s built for the user, not the developer.
You decide exactly what gets stored.
You decide what gets shared (via scoped PATs).
You decide what stays hidden.

This user-owned approach gives far better transparency and explainable AI... the user can always see, audit, and revoke their own memory data.

I’ve also just launched it on Product Hunt as Cortex Memory MCP (Universal Long-Term Memory for AI Agents via MCP):
https://www.producthunt.com/products/cortex-memory-mcp

If you like the idea of a user-owned, portable memory layer (Biological Funnel + HNSW+BM25 hybrid search + Gemini 3 Dreamer pipeline for dedup/conflict resolution), I’d really appreciate an upvote and any quick feedback on the PH page.

Still 100% built with Gemini 3 and fully MCP-native.
The philosophy stays the same: memory should be user-owned and work across every tool (and eventually robots).

Excited for the results whenever they drop!

Try the latest live version here: https://cortexmcp.vercel.app

Old version: https://cortex-protocol.vercel.app