MCP & LLM Tooling — Open-Source Building Blocks

Overview

Beyond building applications with AI, I build the plumbing underneath it — the libraries and Model Context Protocol (MCP) servers that agents plug into. This is a collection of those reusable pieces, each solving a real friction point in the agent stack.

The Building Blocks

LLM Factory — Unified LLM & Parsing Gateway

A modular Python framework that puts one interface in front of many LLM providers (Azure OpenAI, OpenAI, Groq, OpenRouter, LM Studio, Ollama) and document parsers (PDF, DOCX, PPTX). Application code calls a single run_pipeline surface; swapping providers or adding a parser never touches business logic. Published as an installable package.

Graphing MCP Server

An MCP server that gives any AI assistant the ability to generate professional diagrams — architecture, flowchart, ERD, user-flow, process, dashboard, and survey types. Ships with 1,487 cloud/infra icons (AWS, Azure, GCP, Kubernetes), context-aware icon selection, and auto-layout so the model places nodes intelligently instead of by hand.

Vedic MCP Server

A domain MCP server exposing accurate Jyotish (Vedic astrology) calculations to AI assistants — birth charts, panchanga, Vimshottari dasha, and yoga detection — computed on the Swiss Ephemeris. A clean example of wrapping a precise, non-trivial computation library as structured MCP tools, resources, and prompt templates.

The Through-Line

• Protocol-native — these speak MCP, the emerging standard for connecting models to tools and data.
• Provider-agnostic — abstraction over vendors is a recurring theme; nothing here is locked to a single model or cloud.
• Reusable — each is a standalone library/server meant to be dropped into someone else’s agent, not a one-off script.

Highlights

• 3 distinct tooling layers: an LLM gateway, a generative MCP server, and a domain MCP server
• 1,487 icons + 7 diagram types in the graphing server
• 6+ LLM providers unified behind one gateway interface