Add initial README.md for Agent Worflows

agentic-workflows/README.md

+# 🧠 AI-Builder: MCP Server Framework
+
+AI-Builder platforms helps you to build, deploy, and onboard intelligent AI agents using standardized, containerized tools via the **Model Context Protocol (MCP)** and **gRPC**.
+
+> ⚡ **No custom wiring**  
+> 🔧 **Define once, deploy anywhere**  
+> 🚀 **Plug in any tool. Power any agent.**
+
+
+## 📦 What is MCP?
+
+**Model Context Protocol (MCP)** is an open standard for enabling AI systems to access external data sources and tools in a structured, secure, and composable way.
+
+Think of it as **USB-C for AI agents** — a uniform interface to expose:
+
+- Tools (actionable functions)
+- Prompts (reusable LLM interactions)
+- Resources (context/data fetchers)
+
+With MCP, LLMs can interact with real-world services with minimal friction.
+
+
+## 💡 Why FastMCP?
+
+[**FastMCP**](https://github.com/jlowin/fastmcp) is a Python-first framework for building MCP servers. 
+
+Define your tool with a decorator, and FastMCP handles the rest.
+
+---
+
+## 🧬 MCP Agent Flow in AI-Builder
+
+The mcp-servers in AI-Builder addresses the following,
+- Protocol handling (via gRPC)
+- Schema generation (JSON Schema)
+- Tool registration and discovery
+- Docker deployment
+
+```User ➝ LLM ➝ MCP Server ➝ gRPC ➝ Tool ➝ Response ➝ LLM ➝ Output```
+
+1. Schema Discovery
+LLM queries all available tool schemas using SendFunctionSchemas.
+
+2. Tool Selection & Invocation
+LLM sends a ToolRequest with input JSON and target tool.
+
+3. Tool Execution
+The server runs the tool and returns a ToolInvocationResult.
+
+## 🛠️ Server Template Structure
+Each MCP tool lives in a self-contained directory:
+
+<pre> 
+mcp-tool-name/
+├── agent.proto               # Shared proto definition
+├── build_and_deploy.sh      # Docker build + push
+├── Dockerfile               # Container config
+├── mcp_<tool_name>.py       # Tool logic
+├── requirements.txt
+└── server.py                # FastMCP server entry 
+</pre>
+💡 Just change mcp_tool_name, write your function, and deploy.
+
+💡 **To create a new tool:**
+
+1. Copy this folder template and rename it (e.g., `mcp-weather`, `mcp-web-scraper`)
+2. Implement your tool function(s) in `mcp_<tool_name>.py`
+3. Update `build_and_deploy.sh` to reflect your tool's Docker image tag
+4. Run `./build_and_deploy.sh` to build and deploy the container
+
+> ✅ Everything else (proto, server, Dockerfile) is reusable — you only write your tool logic.
+
+
+## 🔌 gRPC Protocol Definition (agent.proto)
+<pre>
+syntax = "proto3";
+
+enum AgentType {
+  UNKNOWN = 0;
+  EVENT = 1;
+  PLACES = 2;
+  WEATHER = 3;
+}
+
+message FunctionSchemasResponse {
+  AgentType agent_type = 1;
+  repeated string function_schemas = 2;
+}
+
+message ToolRequest {
+  AgentType agent_type = 1;
+  string tool_name = 2;
+  string arguments_json = 3;
+}
+
+message ToolInvocationResult {
+  AgentType agent_type = 1;
+  repeated string results = 2;
+}
+
+service ToolService {
+  rpc SendFunctionSchemas (Empty) returns (stream FunctionSchemasResponse);
+  rpc InvokeTool (stream ToolRequest) returns (stream ToolInvocationResult);
+}
+</pre>
+
+## 🚀 Available MCP Servers
+### 🌦️ mcp-weather
+Tool: get_weather_forecast(location: str)
+Provides real-time weather + 24-hour forecast using Open-Meteo API.
+
+Natural language location (via geopy)
+
+Temperature, humidity, wind speed
+
+JSON output with timezone, elevation, etc.
+
+### 🔗 mcp-web-scraper
+Tool: scrape_and_clean(url: str)
+Scrapes and cleans webpage content.
+
+Removes scripts/styles
+
+Extracts readable text and headings
+
+Parses structured data (JSON-LD)
+
+### 📍 mcp-places
+(Coming soon)
+Location-based information and nearby places.
+
+### 🗞️ mcp-news-aggregator
+(Coming soon)
+Aggregates, classifies, and summarizes live news articles.
+
+### 📚 mcp-wiki-knowledge-extractor
+(Coming soon)
+Extracts structured summaries and linked concepts from Wikipedia.
+
+
+## 🧪 Sample Pipelines
+You can combine these MCP tools with LLMs to build agent workflows such as:
+
+| Pipeline               | Tools Used                                      | Description                                                       |
+|------------------------|--------------------------------------------------|-------------------------------------------------------------------|
+| 🗓️ Event Planner Agent | `mcp-web-scraper`, `mcp-weather`, `mcp-places`   | Recommends weather and nearby venues for an event                |
+| ✈️ Travel Advisor Agent| `mcp-weather`, `mcp-wiki`, `mcp-web-scraper`     | Gives travel suggestions, weather tips, and wiki facts           |
+| 🗞️ News Summarizer Bot | `mcp-news-aggregator`, `mcp-web-scraper`         | Provides summarized, cleaned news from live sources              |
+| 🧠 Knowledge Builder    | `mcp-wiki`, `mcp-web-scraper`                    | Builds structured info graphs from natural queries               |
+
+
+## 📦 Getting Started
+1. Clone or fork the template
+git clone 
+
+2. Navigate to tool directory
+cd 
+
+3. Edit `mcp_<tool_name>.py` and define your tool logic
+
+4. Build & deploy container - change the docker tag accordingly
+chmod +x build_and_deploy.sh
+./build_and_deploy.sh
+
+## 📈 Future Roadmap
+- Agent memory & conversation history
+- Dynamic tool discovery
+- Multi-tool support in one server
+- Full support for Prompt, Resource, and Schema features of FastMCP
+
+## 🤝 Contribute
+Want to add your own tool? Fork this repo and follow the structure to create mcp-<your-tool-name>. PRs welcome!
+
+## 📫 Contact
+For feedback or collaboration, reach out via issues or Discussions.
+
+## 🧭 Explore. Extend. Build.
+We now support:
+
+✅ 5 MCP servers  
+✅ 3 LLMs with tool-calling enabled  
+✅ Standardized onboarding in AI-Builder
+
+Build your own agent pipelines — with no boilerplate, no hassle.
+> Write your tool once.  
+Use it across all agents.  
+Deploy in minutes with AI-Builder.  
\ No newline at end of file