A powerful persistent memory enhancement for AI agents that enables long-term context retention and semantic retrieval across multiple sessions.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install cortex-mem-mcp
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install cortex-mem-mcp using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Cortex Memory MCP?
cortex-mem-mcp provides a robust framework for adding long-term memory to your AI workflows. By integrating this skill with your agent, you can store user preferences, project context, and past interactions that persist beyond single chat sessions. It utilizes semantic search and a tiered data structure to ensure that the AI recalls the most relevant information without overwhelming its context window.
This implementation of Openclaw Skills bridges the gap between ephemeral chat interactions and truly personalized AI assistance. It transforms standard agents into context-aware assistants that grow more useful the more you interact with them, leveraging a vector database for high-performance retrieval.
Cortex Memory MCP Use Cases
- Storing and recalling specific user preferences and settings across different projects and sessions.
- Maintaining project-specific knowledge bases that automatically expand as the conversation progresses.
- Persisting deep conversation context for long-running software development or research tasks.
- Tracking historical user-agent interactions to improve personalization and reduce repetitive instructions.
- Performing semantic searches across historical data using natural language queries to find forgotten details.
How Cortex Memory MCP Works
- The AI agent captures relevant information during a session and uses the store tool to place it in the memory pipeline.
- Upon reaching a message threshold or receiving a manual command, the commit tool triggers memory extraction.
- The system generates tiered layers (L0 Abstract, L1 Overview, and L2 Full Content) to balance detail with token efficiency.
- Information is transformed into vector embeddings and indexed in a Qdrant database for semantic retrieval.
- When context is needed later, the agent uses search or recall tools to fetch the most relevant memory layers based on the current prompt.
Cortex Memory MCP Setup
First, verify if the binary is available or install it via cargo:
cargo install cortex-mem-mcp
You must have a vector database running. The easiest way is using Docker:
docker run -d -p 6333:6333 qdrant/qdrant
Create a config.toml specifying your LLM API keys and database URL. Finally, add the server to your MCP client (such as Claude Desktop or Cursor). Integrating this into your Openclaw Skills workflow requires pointing the client to your config path using the --config flag.
Cortex Memory MCP Data Schema & Taxonomy
The skill organizes data using a structured URI scheme and a tiered content model to optimize retrieval speed and context window usage:
| URI Pattern | Purpose |
|---|---|
cortex://session/{id}/conversation.md |
Session-specific history and logs |
cortex://user/{id}/preferences/{topic}.md |
User-specific settings and style guides |
cortex://user/{id}/memories/{id}.md |
General persistent knowledge snippets |
Memory Layers:
- L0 (Abstract): ~100 tokens. Used for quick relevance checking.
- L1 (Overview): ~2000 tokens. Provides a condensed understanding of the core information.
- L2 (Content): Full original content. The complete source document.
Cortex Memory MCP Advanced Features
- Tiered access model (L0/L1/L2) allowing the agent to preview relevance before loading full documents.
- Automatic memory processing triggered by message count thresholds or periods of inactivity.
- Multi-tenant support enabling complete memory isolation between different users or projects via tenant IDs.
- Smart exploration tools that combine hierarchical URI browsing with semantic search capabilities.
- High-performance vector indexing using Qdrant for millisecond retrieval across thousands of memory entries.
SKILL.md
Loading
Related Openclaw Skills
A-Share Financial Analysis
A professional-grade financial diagnostic tool that generates multi-dimensional analysis reports for A-share listed companies via remote AI agents.
LocalMiniDrama Assistant
An end-to-end AI assistant for generating short dramas from natural language scripts to final video synthesis.
MyReels Storyboard Design Tool
A professional-grade storyboard planning tool designed to streamline character consistency and shot design for short-form video production.
Audio SRT Workflow
A comprehensive toolkit for generating, aligning, and validating SRT subtitles from audio files using faster-whisper and ffmpeg.
pinkr-crm
A Python-based CLI tool for AI agents to interact seamlessly with Pinkr CRM backend APIs for customer data management.
AI Companion Market Research Automator
A fully automated agent skill for generating comprehensive market research reports on the global AI companion software sector.
Featured*
