UseCortex MCP is a Model Context Protocol (MCP) server designed to give AI coding agents persistent memory. It lets agents read the user's stored knowledge and write new findings back, all through an encrypted endpoint.
Purpose
This tool is aimed at AI coding scenarios that need to maintain context across sessions. Users can describe their coding standards, architectural decisions, customer requirements, and more through chat, and UseCortex automatically extracts and organizes this information into structured topics that agents can call on at any time. This avoids context window limits, repeated copy-pasting, and outdated documentation.
Main capabilities
- Bidirectional knowledge flow: agents can both read the context needed for a project and write new findings back to the knowledge base.
- Compatible with any MCP client: supports all AI coding agents that conform to the Model Context Protocol.
- Encryption at rest: uses AES-256 encryption with a separate key for each user.
- No vendor lock-in: user knowledge can be used across different AI tools.
- Organized by topic: knowledge is automatically categorized by topic instead of being piled together.
- Provides multiple tools: including query_knowledge (ask questions in natural language and get AI answers based on stored knowledge), list_topics (list all knowledge topics), write_knowledge (write new knowledge), get_topic (get all knowledge entries for a specified topic), search_knowledge (search knowledge entries by keyword); under the Memory plan it also provides capture_session (store an AI session summary in persistent memory), recall_memory (use AI to search all captured session memories), list_sessions (list recent captured session memories).
How to use
- Generate an API key: register at usecortex.net, go to Settings → API Keys → Generate, and copy the key immediately after generating it (it will not be shown again).
- Install: run an installation command in the terminal, replacing YOUR_API_KEY with the key generated in the first step.
- Restart the AI coding agent, and the UseCortex tools will be available. If you prefer manual configuration, you can add the corresponding configuration item to the MCP client's configuration file, also replacing YOUR_API_KEY. The prerequisite for use is having a free UseCortex account and an MCP-compatible AI coding agent.
Applicable scenarios
Suitable for developers who need AI coding agents to remember coding standards, architectural decisions, customer requirements, project context, and other information over the long term. Any AI coding tool that supports MCP can connect.
Limitations and security
All communication is encrypted via HTTPS; knowledge at rest is encrypted with AES-256; each user has a separate encryption key, and the official statement says that even they cannot read user data; API keys are divided by user and can be revoked instantly. Currently no detailed pricing information is provided; it only mentions that plans can be managed at usecortex.net. Other limitations are not yet provided.
