13. Building Custom MCP Servers from Scratch in TypeScript / Node.js and Python • AI Autonomous Agents, Tool Calling & MCP (LangGraph) • Rishabh Ki Kaksha

AI Autonomous Agents, Tool Calling & MCP (LangGraph) › 13. Building Custom MCP Servers from Scratch in TypeScript / Node.js and Python

55 mins

← Prev Next →

Course Syllabus

Module 1: Agentic AI Foundations, Reasoning Loops & Tool Calling

01. What are Autonomous AI Agents? The Shift from Passive Chatbots to Action-Taking Systems 02. Cognitive Architectures: ReAct (Reason + Act), Plan-and-Solve & Reflexion Memory Loops 03. Function Calling & Tool Execution: JSON Schema Contracts, Dynamic Dispatch & Error Recovery

Part 1: Agentic Patterns, Graphs & Protocols

Module 2: LangGraph Architecture: StateGraphs, Nodes, Edges & Memory Checkpointing

04. LangGraph Core: Why DAGs Fail and Why Cyclical Graph Architecture is Mandatory for Agents 05. StateGraph Architecture: TypedDict State Schemas, State Reducers & Channel Updates 06. Conditional Edges: Routing Agent Actions dynamically based on LLM decisions 07. Persistence & Memory: SQLite / PostgreSQL Checkpointers for Resumable Agent Conversations

Module 3: Multi-Agent Collaboration & Supervisor Patterns

08. Multi-Agent Systems (MAS): The Supervisor Pattern (Central Coordinator Orchestrating Specialized Agents) 09. Peer-to-Peer Multi-Agent Communication & Consensus Protocols (Researcher, Coder, Critic, Reviewer) 10. CrewAI vs AutoGen vs LangGraph: Architectural Benchmarks for Enterprise Production

Module 4: Model Context Protocol (MCP) Architecture & Custom Server Building

11. Model Context Protocol (MCP) Deep Dive: The Universal Open Standard for AI Tool Integration (Anthropic Spec) 12. MCP Architecture: Transports (stdio vs SSE), Resources, Prompts & Tool Declarations 13. Building Custom MCP Servers from Scratch in TypeScript / Node.js and Python 14. Connecting MCP Servers to Antigravity, Claude Desktop & LangGraph AI Pipelines

Module 5: Human-in-the-Loop, Time-Travel Debugging & Autonomous Workflows

15. Human-in-the-Loop (HITL): Interrupts, Approval Gates & Manual State Override before Tool Execution 16. Time-Travel Debugging: Rewinding Agent State, Modifying Memory & Forking Execution Branches 17. Sandboxed Tool Execution: Running Untrusted Code safely in Docker Containers & E2B Sandboxes 18. Production Case Study: Building an Autonomous Full-Stack Software Engineer AI Agent from Scratch

Interactive Lesson Studio

# 13. Building Custom MCP Servers from Scratch in TypeScript / Node.js and Python

> AI Autonomous Agents & LangGraph Notes by Rishabh - Topic 13

---

### 1. Complete Custom TypeScript MCP Server from Scratch 🛠️

```typescript
import { Server } from "@modelcontextprotocol/sdk/server/index.js";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
import { CallToolRequestSchema, ListToolsRequestSchema } from "@modelcontextprotocol/sdk/types.js";
import { z } from "zod";

const server = new Server(
    { name: "production-deployer", version: "1.0.0" },
    { capabilities: { tools: {} } }
);

// 1. Declare Available Tools:
server.setRequestHandler(ListToolsRequestSchema, async () => {
    return {
        tools: [
            {
                name: "deploy_service",
                description: "Deploys service to production VPS",
                inputSchema: {
                    type: "object",
                    properties: {
                        serviceName: { type: "string" },
                        branch: { type: "string", default: "main" }
                    },
                    required: ["serviceName"]
                }
            }
        ]
    };
});

// 2. Handle Tool Invocation:
server.setRequestHandler(CallToolRequestSchema, async (request) => {
    if (request.params.name === "deploy_service") {
        const { serviceName, branch } = request.params.arguments as any;
        // Business logic execution:
        const result = `🚀 Successfully deployed ${serviceName} from branch ${branch || 'main'}!`;
        return {
            content: [{ type: "text", text: result }]
        };
    }
    throw new Error(`Tool not found: ${request.params.name}`);
});

// 3. Connect via Standard I/O:
async function run() {
    const transport = new StdioServerTransport();
    await server.connect(transport);
}
run();
```

---

### 2. Interview Ke 3 Sawal Is Topic Se

1. **Zod schemas TypeScript MCP servers me input validation kaise streamline karte hain?** -> `z.object({...}).parse(args)` se runtime type validation aur automated JSON Schema generation provide karna.
2. **Error handling MCP tool execution me kaise communicate hoti hai?** -> Response me `isError: true` flag set karke error message content text me pass karna taaki LLM tool failure dekh sake aur self-heal kare.
3. **FastMCP Python framework boilerplate code kaise reduce karta hai?** -> `@mcp.tool()` decorator use karke type-hints aur docstrings se automatically tool schema aur dispatch generate karna.

---
**Next Topic:** 14. Connecting MCP Servers to Antigravity, Claude Desktop & LangGraph AI Pipelines

Loading lesson content...

Python 3.12

Full IDE

Live Execution Terminal

Click "Execute Live" to run code.
                

Lesson Mastery AI Quiz

AI Doubt Solver

CERTIFICATE OF COMPLETION

Rishabh Gehlot

PHP 8+ & Laravel 11 Master Track