Fozikio

Fozikio / cortex-engine

Public

Cognitive engine for AI agents — semantic memory, observations, embeddings, dream consolidation. Cloud Run service + MCP tools.

15
3
100% credibility
Found Mar 15, 2026 at 11 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

Cortex-engine adds persistent, semantic memory to AI agents so they retain knowledge, beliefs, and experiences across sessions and models.

How It Works

1
🔍 Discover persistent memory for your AI helper

You hear about a simple way to make your AI assistant remember things forever, even after it restarts.

2
📁 Set up in your project folder

In your work folder, you run a quick setup command that adds memory tools and safety guides automatically.

3
🚀 Launch the memory service

With one click, you start the memory helper that runs quietly in the background, ready for your AI.

4
💭 Tell your AI to observe and query

Your AI starts noticing facts with 'observe' and pulling past knowledge with 'query', building experience over time.

5
😴 Run the dream cycle

You trigger a 'dream' to consolidate short notes into strong, lasting memories, like sleep for your AI.

6
🧠 Watch your AI grow smarter

Over sessions, your AI recalls beliefs, connections, and lessons without reminders, developing real personality.

🎉 Your AI has true long-term memory

Now your assistant remembers everything from past work, acts consistently, and feels truly intelligent.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 11 to 15 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is cortex-engine?

Cortex-engine is a TypeScript cognitive engine that gives AI agents persistent semantic memory across sessions, solving the "amnesiac agent" problem where models forget everything after one chat. It stores observations, beliefs, and hypotheses as interconnected nodes with embeddings, belief updates, dream consolidation for long-term compression, and FSRS spaced repetition to keep retrieval fresh. Deploy as a Cloud Run service or local SQLite app, exposing 25 cognitive tools over the MCP protocol for clients like Claude Code or Cursor.

Why is it gaining traction?

Unlike prompt-only agents, it builds emergent personality and expertise from accumulated experience—no more repeating "you care about distributed systems." The hook is dead-simple setup: npx fozikio init creates a workspace with Reflex safety rules, hooks, and multi-agent namespaces; agents just call query(), observe(), or dream() via MCP. LLM-agnostic with pluggable Ollama/Vertex providers, plus plugins for threads, journals, and evolution tracking.

Who should use this?

AI agent builders in Claude Code, Cursor, or Windsurf who need stateful cognition for coding projects, research workflows, or trading bots. Ideal for devs engineering cognitive architectures where agents ground decisions in past observations, like reviewing codebases or evolving strategies over sessions.

Verdict

Early alpha with 11 stars and 1.0% credibility—test coverage shines via e2e validation scripts, but expect rough edges in production. Grab it for cognitive engineering experiments if you're prototyping persistent agents; skip for mission-critical unless you contribute.

(187 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.