Personal Task MCP Server
A local task-management server exploring how AI applications connect to tools and data through MCP.
- Python
- MCP
- Gemini
- Google ADK
- McpToolset
- JSON
- stdio
Project overview
This project was built as a hands-on learning project to understand the Model Context Protocol (MCP) from the ground up. I developed a local Python-based MCP server that provides task-management capabilities and connected it with custom MCP clients, Gemini, and a Google ADK agent. The project demonstrates the complete flow from a natural-language user request to an AI model, MCP client, MCP server, MCP tool, local task data, and the returned result. Local communication is handled using stdio, while tasks are stored in a simple JSON file to keep the focus on learning MCP rather than database engineering.
Preview
Project preview
Key features
- Local Python-based MCP server for task management.
- Task tools: add, list, complete, delete and search.
- MCP resources (tasks://all) and prompts (daily_task_review).
- Custom MCP client with tool discovery and invocation.
- Gemini and Google ADK (McpToolset) integrations.
- stdio-based local communication with JSON storage.
Technologies
- Python
- MCP
- Gemini
- Google ADK
- McpToolset
- JSON
- stdio
What I learned
I learned how MCP acts as a standard connection layer between AI applications and external tools, including tool discovery, invocation, resources, prompts and local stdio communication.