Memvex (formerly Cortex)
Your agents don’t know you. Memvex fixes that.
Memvex is a local-first MCP server that provides a shared runtime for all your AI agents. Instead of each agent starting from scratch, Memvex gives them access to:
- Identity: Your preferences, coding style, and project context.
- Memory: A shared knowledge graph that persists across sessions and tools.
- Guardrails: Rules to prevent agents from doing things they shouldn’t (like spending too much money or deleting production DBs).
How it works
- Install & Init:
npx memvex init - Configure: Edit
memvex.yaml - Connect: Add Memvex as an MCP server to Claude, Cursor, or any MCP-compatible tool.
Designed to be local-first and privacy-focused.
Key Features
- Shared identity: Define your coding style, preferences, and project context once
- Persistent memory: Knowledge graph that remembers information across all sessions
- Action guardrails: Set rules to prevent dangerous or costly operations
- MCP compatible: Works with Claude, Cursor, Windsurf, and any MCP-enabled tool
- Local-first: All data stays on your machine for maximum privacy
Use Cases
- Consistent coding: Maintain consistent coding style across all AI assistants
- Project context: Share project-specific knowledge with every agent you use
- Safety: Prevent agents from making destructive changes to production systems
- Continuity: Pick up where you left off across different AI tools and sessions
- Team collaboration: Share identity and memory configurations with your team
Benefits
Memvex solves the fundamental problem of AI agents having no memory or context between sessions. By providing a shared runtime, it transforms isolated AI interactions into a cohesive, intelligent system that learns and adapts. The local-first architecture ensures your data never leaves your machine, addressing privacy concerns that come with cloud-based AI services. Action guardrails provide peace of mind by preventing costly mistakes. As an open-source project, it’s transparent and extensible for developers who want to customize their AI agent infrastructure.