Shared external memory
Cortex stores semantic memory outside the model so different AIs and agents can draw from the same knowledge base instead of keeping separate notes.
Cortex
Cortex provides semantic memory infrastructure for AI assistants, sharing persistent context, decisions, and sources with selective memory storage and contradiction tracking.
Semantic memory infrastructure for AI assistants
Cortex is a semantic memory infrastructure product from SKYNETLAB. It is designed as an external memory layer that multiple AIs can share, rather than a model embedded inside a single assistant. The site describes it as patent pending, in production, and model agnostic.
Its main job is to preserve decisions, context, and sources across conversations so an assistant can keep working with the same knowledge over time. The product is also documented as an MCP connector for Claude, where it creates a personal memory space on first sign-in and lets the assistant save, search, synthesize, and retrieve memories in later sessions.
Core capabilities
Selective memory filtering
The system rejects roughly eight out of ten items it is asked to store, keeping only information that adds value and flagging contradictions when they appear.
Contradiction tracking
When a new note conflicts with earlier memory, Cortex keeps both versions and marks the contradiction instead of silently overwriting one of them.
MCP-based connectivity
Through MCP, connected assistants can reach the shared memory and, from there, also access the user's files, email, and archives as part of the same workflow.
Structured memory actions
The Claude connector exposes tools for semantic search, memory detail lookup, narrative recall, source review, saving, and confirmed deletion.
Self-checking and source-aware memory
The product says it compares predictions with outcomes and distinguishes verified facts from hypotheses, so the memory is not just a passive archive.
Practical ways to use Cortex
Project memory that survives new chats
Keep a stable record of project decisions, supplier choices, and working assumptions so the next conversation starts with the same context.
Model switching without losing context
Use the shared memory with different assistants or models without rebuilding the same background information each time you switch tools.
Reviewing decisions and evidence
Let Claude retrieve prior memories, show sources, and surface conflicts when you need to check what was said and why it was stored.
Filtering what deserves to stay
Save durable facts during an ongoing discussion while leaving transient or duplicate notes out of the memory store.
Controlled cleanup of outdated memories
Use the connector's confirm-before-delete flow when you need to remove a memory while keeping the rest of the account intact.
Pros and Cons
Pros
- Works as an external memory layer that can persist across conversations and model changes.
- Supports shared access through MCP, so connected assistants can use the same memory instead of fragmenting context.
- Tracks contradictions rather than hiding them, which can make memory review more transparent.
- Provides specific actions for search, retrieval, synthesis, saving, and deletion through the Claude connector.
- Includes a free 30-day period in the Claude guide before paid use begins.
Cons
- The public pricing page returns a 404, so the current pricing structure is not fully visible from the main site.
- The strongest setup documentation is for Claude; support for other MCP clients is mentioned, but the source provides fewer setup details for those clients.
- The product is oriented around persistent memory and quality checks, which may be more than a simple note store for users who only need basic chat history.
FAQ
How does Cortex work with Claude?
Cortex is meant to be connected as an external semantic memory, so Claude can save and retrieve decisions, context, and sources across later sessions. The Claude guide says no special commands are needed; you talk to Claude normally and it chooses the right memory action when needed.
What is required to set up Cortex with Claude?
The Claude setup guide says you need a Claude plan that supports custom connectors. In the guide, Cortex is added through Claude settings by pasting a connector address, then signing in with Google or email to create your personal memory space.
Is Cortex memory shared between users?
The documentation says each user gets a personal, isolated memory space. Memories are personal, exportable, and deletable on request, and disconnecting Cortex from Claude does not delete the memories from the Cortex account.
Does Cortex have pricing or a free trial?
The source material does not show a current public pricing page. The Claude guide says there is a 30-day free period and then pricing starts from 0,99€ per month.
What can the Cortex connector do inside Claude?
The Claude guide documents six connector tools: Search memories, Fetch memory, Recall & synthesize, Show conflicts, Save memory, and Forget memory. The write and delete actions include quality checks or explicit confirmation.
Quick Facts
- Category
- AI memory infrastructure
- Primary integration
- Claude via remote MCP connector
- Platform notes
- Model-agnostic; other MCP clients are referenced
- Memory behavior
- Selective storage with contradiction tracking
- Pricing signal
- 30 days free, then from 0,99€ per month
- Website
- skynetlab-cortex.com
Cortex Alternatives
Falconer
Falconer is an AI-powered company brain that keeps internal docs accurate, searchable, and up to date by syncing with GitHub, Slack, and Linear.
ByteAsk
ByteAsk is a terminal-first AI coding agent for C and C++ that edits repos and verifies changes with compilers, debuggers, sanitizers, and tests.
Goldfish
Goldfish is an AI memory app for macOS that helps you reply, write, summarize, and continue with context already on your Mac. It stores memory locally and writes in your tone.
Lasso
Lasso is an ecommerce product data platform for enriching catalog records, processing supplier files, generating product content, and tracking competitors. It includes a web app, REST API, SDK, and MCP server.
CreateOS Sandbox
CreateOS Sandbox is an isolated compute environment for running code and agent workloads in Firecracker micro-VMs with private networking and SDK, CLI, or MCP control.
ArtDeck
ArtDeck is a visual reference board app for iPhone, iPad, and Mac. Collect mixed references on one canvas and study them with ToolBox lenses. Offline, optional iCloud sync, one-time purchase.