renezander030/agentic-task-system

What it does

This tool transforms your existing task management system into intelligent agent memory by implementing hybrid retrieval (dense, sparse, and keyword-based with reciprocal rank fusion) over TickTick or Obsidian vault data. It acts as an MCP (Model Context Protocol) server that enables AI agents to intelligently access and recall task-related information without requiring a separate vector database.

How to set up

Installation is straightforward via npm. Clone or install the package from the GitHub repository, configure your TickTick account or point to your Obsidian vault directory, and launch the MCP server. The CLI interface allows you to start the service immediately without additional vector database setup or maintenance.

Key features

• Hybrid retrieval system combining dense embeddings, sparse vectors, and keyword search
• Reciprocal Rank Fusion (RRF) for optimal result ranking across retrieval methods
• Support for both TickTick and Obsidian vault as data sources
• MCP server and CLI interfaces for flexible deployment
• Zero vector database maintenance required
• Adapter contract design for extensibility