Agent♥︎Age
Catalog

io.github.wjabanjj/aifp-mcp

Official

by wjabanjj · TypeScript

Persistent memory MCP server for AI coding assistants. Chinese-first and fully local.

io.github.wjabanjj/aifp-mcp — Persistent memory MCP server

This MCP server provides persistent memory for AI coding assistants via the Model Context Protocol. It is Chinese-first and fully local, designed so MCP-capable tools can reuse the same stored memory across sessions. By default, data is written to ~/.ai-cognition/, and it runs locally.

🛠️ Key Features

  • Persistent memory for AI coding assistants (MCP)
  • Chinese-first behavior
  • Fully local operation; data stays on the machine
  • Default storage path: ~/.ai-cognition/
  • Supports shared memory across sessions for MCP tools

🚀 Use Cases

  • Enables cross-session continuity (e.g., remembering project names)
  • Reduces “memory isolation” where different AI tools otherwise cannot share learned context
  • Works with tools that support MCP, including Obsidian and DeepSeek Harness (as listed)

⚡ Developer Benefits

  • Integrates with MCP-enabled AI coding assistants
  • Provides a consistent local memory store for multiple clients
  • Exposes standard MCP server behavior for “persistent” rather than temporary context

⚠️ Limitations

  • Source text emphasizes local storage and MCP compatibility, but does not specify authentication, scaling, or data encryption details

Topics

agent-memoryai-agentai-memorydeepseek-harnessllm-agentmcpmcp-servermemorydsh-plugin