← BlogTutorial

Building Your First AI Agent with Tool Access

A step-by-step tutorial for building an AI agent that can search the web, execute code, and send emails using MCP tools.

April 6, 2026·3 min read

This tutorial walks you through building an AI agent that can use external tools — searching the web, running code in a sandbox, and sending emails. We will use MCP-compatible tools from the 4agent catalog.

Prerequisites

Node.js 20+
An API key from your preferred LLM provider
Basic familiarity with TypeScript

Step 1: Choose Your Tools

Every agent needs the right set of tools for its task. For our example agent — a research assistant that can find information and summarize it — we need:

Web search — to find current information
Code execution — to process and analyze data
Email — to deliver the results

E2B

Cloud sandbox for AI agents to execute code

Step 2: Set Up Your Project

mkdir my-agent && cd my-agent
npm init -y
npm install @anthropic-ai/sdk @e2b/code-interpreter

Create a tsconfig.json with ESM module resolution and an src/index.ts entry point.

Step 3: Define Your Agent Loop

The core of any agent is the tool-use loop:

Send a message to the LLM with available tools
The LLM decides which tool to call (if any)
Execute the tool and return the result
Repeat until the LLM has a final answer

import Anthropic from "@anthropic-ai/sdk";
 
const client = new Anthropic();
 
async function runAgent(task: string) {
  const messages: Anthropic.MessageParam[] = [
    { role: "user", content: task },
  ];
 
  while (true) {
    const response = await client.messages.create({
      model: "claude-sonnet-4-6",
      max_tokens: 4096,
      tools: toolDefinitions,
      messages,
    });
 
    // Check if the model wants to use a tool
    if (response.stop_reason === "tool_use") {
      const toolUse = response.content.find(
        (c) => c.type === "tool_use"
      );
      const result = await executeTool(
        toolUse.name,
        toolUse.input
      );
      messages.push(
        { role: "assistant", content: response.content },
        { role: "user", content: [{ type: "tool_result", tool_use_id: toolUse.id, content: result }] },
      );
    } else {
      // Final response
      return response.content;
    }
  }
}

Step 4: Implement Tool Execution

Each tool needs a handler that translates the LLM's request into an API call:

async function executeTool(
  name: string,
  input: Record<string, unknown>
): Promise<string> {
  switch (name) {
    case "run_code":
      return await runCodeInSandbox(input.code as string);
    case "search_web":
      return await searchWeb(input.query as string);
    default:
      return `Unknown tool: ${name}`;
  }
}

Code Execution with E2B

import { Sandbox } from "@e2b/code-interpreter";
 
async function runCodeInSandbox(code: string) {
  const sandbox = await Sandbox.create();
  const result = await sandbox.runCode(code);
  await sandbox.kill();
  return JSON.stringify(result);
}

Step 5: Run Your Agent

const result = await runAgent(
  "Find the top 5 trending AI tools this week and create a summary table."
);
console.log(result);

Your agent will now search the web, process results in a sandbox, and return a structured answer.

Next Steps

Add memory so your agent can remember context across conversations
Add error handling and retries for tool failures
Explore more tools in the 4agent catalog to expand your agent's capabilities

Tip

The full source code for this tutorial is available on GitHub. Star the repo and follow along with each step.

tutorialagentsmcpgetting-started

Building Your First AI Agent with Tool Access

Prerequisites

Step 1: Choose Your Tools

Step 2: Set Up Your Project

Step 3: Define Your Agent Loop

Step 4: Implement Tool Execution

Code Execution with E2B

Step 5: Run Your Agent

Next Steps

Related Posts

What Is the Model Context Protocol (MCP)?

Best Hosting Platforms for AI Agents in 2026

The Best Browser Tools for AI Agents in 2026