Mengcp [re: me-nge-cap]

Random tools coded while learning and exploring MCP (Model Context Protocol) for building AI agents.

Tools

Note

All tools are designed for my personal use, so they may not work as expected for you. Feel free to modify the code to suit your needs.

Code Editing Agent

Contains simple tools for code editing by AI agent using Claude (currently). It's based on the tutorial from AmpCode with slight modifications to the code structure and file organization for improved tidiness and performance.

• More modular code with organized file structure
• Performance optimizations in the code with map and in-memory cache for the definition setting

Feel free to explore and modify the code for your own purposes!

Poke Agent

Inspired by allenthomas technical walkthrough to explain what and how MCP works in simple terms.

Finance Agent

This agent is designed to fetch stock data from Yahoo Finance and EODHD for financial analysis. Inspired by this micro cap experiment.

Development

Prerequisites

• Go 1.23 or later
• Anthropic API key

Local Setup

Caution

Beware that this project is still in its early stages and may not be fully functional or stable yet. Use at your own risk. Especially the function where the agent needs to read and write files in your local machine.

1. Clone the repository
2. Install dependencies using go mod tidy
3. Create .env file by copy the content from .env.example and adjust the value for each variable, especially ANTHROPIC_API_KEY with your own API key from Anthropic. EODHD_API_KEY is optional, you can use it if you want to use EODHD as a fallback for Yahoo Finance.
4. Run the application using go run main.go

   Ensure you have the necessary permissions to run the application, especially if you're using a Mac

Diagram

Check out the diagram for the visual explanation.

Demo

See the demo for a quick overview of the tools in action.

Todo

• add tools for interacting with hosted actual budget
• add tools for interacting with local logseq graph
• trading analysis tools using historical data from Yahoo Finance and EODHD againts the current gotrade portfolio
  • use tavily for investor sentiment scraping news
  • store news as vector in vector database, milvus/qdrant/pgvector?