The full stack for memory-aware agents
From persistent memory to cognitive plugins to multi-agent workspaces — everything your agent needs to think, not just respond.
Memory in ten lines
Give your agent persistent, semantic memory. No vector DB setup, no embedding pipelines — cortex-engine handles it.
import { CortexEngine } from 'cortex-engine'; const cortex = new CortexEngine({ provider: 'sqlite', embeddings: 'ollama',}); // Your agent observes somethingawait cortex.observe('User prefers TypeScript over Python'); // Later, it recalls relevant contextconst memories = await cortex.query('language preferences');// → [{ content: 'User prefers TypeScript...', salience: 0.92 }]Up and running in one command
cortex-engine handles storage, embeddings, and MCP tooling. You focus on your agent.
Install
npm install cortex-engine, then npx fozikio init. Local SQLite storage, Ollama embeddings, default config. No accounts needed.
Configure
Choose your storage provider (Firestore, SQLite), embedding engine (Ollama, Vertex, OpenAI), and install plugins. Or just use the defaults.
Build
observe(), query(), validate(), evolve() — 17 core MCP tools plus plugin tools. Your agent remembers everything.
Built in the open
Everything is MIT licensed. Read it, fork it, contribute. 10 stars on GitHub is worth more than 0 sales on Gumroad.