memory

This Memory MCP Server provides long-term memory capabilities for LLMs running in Cursor by connecting to HPKV's memory platform. It enables the AI assistant to remember project structures, coding conventions, user preferences, prior explanations and decisions, and to avoid revisiting failed approaches. The server is accessed via the Cursor MCP integration and the HPKV memory service at the configured URL. To use it, add the Memory MCP Server to your Cursor setup (via mcp.json) and authenticate with your HPKV account when prompted. Once enabled, you can attach memory rules to your projects so the AI can persistently recall relevant information across sessions.

Tools and capabilities include storing important context (e.g., project conventions, preferred patterns, and prior solutions) and retrieving that context when needed during coding sessions. The system is designed to reduce hallucinations by providing persistent references to past decisions and memory of prior attempts, making AI coding assistants more reliable over time.

Prerequisites:

• Access to Cursor IDE with MCP support
• An HPKV account and an API key created in the HPKV dashboard
• Internet access to reach the HPKV memory service

Setup steps:

1. Create or edit your mcp.json in your project to include the HPKV Memory MCP Server:

{
  "mcpServers": {
    "hpkv-memory-server": {
      "url": "https://memory.hpkv.io/sse"
    }
  }
}

2. In Cursor, enable the Memory MCP Server you added. You will be prompted to login to your HPKV account and select the API key you generated.

3. Add the memory cursor rule to your project and set the rule type to Always to ensure memory is consulted for relevant interactions.

4. (Optional) Integrate Memory with Claude Code by following the repository's guidance:

claude mcp add -s user -t sse hpkv-memory-server https://memory.hpkv.io/sse

5) Verify that the MCP server is reachable and that authentication succeeds. If you encounter issues, refer to the troubleshooting notes below.

Notes and tips:

• This MCP server relies on HPKV's memory service. Ensure your HPKV API key is valid and that the memory endpoint (https://memory.hpkv.io/sse) is accessible from your network.
• If authentication falters, you may need to disable the MCP server in Cursor, terminate mcp-remote processes, clear cached credentials (e.g., rm -rf ~/.mcp-auth), then re-enable and re-authenticate.
• Use the provided memory-related rules (e.g., Always) to ensure long-term memory is considered during assistant responses.
• Typical environment considerations include network latency to HPKV and the size of memories being stored; avoid storing excessively large memories if not necessary.
• Refer to HPKV documentation for any feature-specific constraints or rate limits.