Getting started

Cortex is a memory layer for AI agents. It stores memories as linked pages and retrieves with vector search plus link-expansion, so multi-hop facts come back without over-fetching a large top-k — ask "who directed Inception?" and get back not just that page, but the linked page about where the director was born, because the two pages share an entity.

There are two ways to run it, with the same method names and result shapes:

Hosted — CortexClient

Talk to a running Cortex server (the hosted one, or your own). No local storage, no models to download — an API key and pip install cortexlayer.

Embedded — Memory

The engine runs in your own process. No server, no API key, nothing leaves your machine. pip install "cortexlayer[local]".

Quickstart: hosted

pip install cortexlayer

import os
from cortexlayer import CortexClient

client = CortexClient(api_key=os.getenv("CORTEX_API_KEY", "sk-..."))

client.add("I moved to Lisbon in March.")

results = client.search("Where does Alice live?", limit=5)
print(results)

Create a key in the Cortex web app (Keys). Full reference: Hosted / CortexClient.

Quickstart: embedded

pip install "cortexlayer[local]"

from cortexlayer import Memory

m = Memory("./cortex-data")

m.add("Christopher Nolan directed Inception.", user_id="alice")
m.add("Christopher Nolan was born in London in 1970.", user_id="alice")
m.relink(user_id="alice")  # batch linking pass — run after adding several

for r in m.search("Who directed Inception?", user_id="alice", limit=1):
    print(r.title, "-", r.via)

# Christopher Nolan directed Inception. - direct
# Christopher Nolan was born in London in 1970. - link   <- pulled in via the shared entity