How to install the Standard MCP Server Part - 1

💡Introduction to the MCP Blog Series

Welcome to our MCP Installation & Customization Blog Series — your step-by-step guide to installing, configuring, and extending MCP servers!📘

In this three-part series, we’ll walk you through everything from setting up an official MCP server to building and integrating custom tools that communicate with REST APIs. Whether you’re a beginner just getting started or an advanced user looking to extend your server’s capabilities, this series has got you covered.

Here’s what you can expect:

🧱 Installing the Official MCP Server

In our first blog, we’ll guide you through the process of installing the official MCP server. You’ll learn how to:

• Prepare your environment

• Download and configure the official server

• Start your first MCP instance

By the end, you’ll have a fully functioning official MCP server ready to serve to your queries.

What is MCP (Model Context Protocol)? – Explained Simply

MCP (Model Context Protocol) is basically an open standard that enables AI agents and Large language Models(LLM) to interact with external systems including salesforce using Natural language instead of traditional API systems.

In the salesforce context, It acts as a universal connector allowing external system and AI agents to access salesforce data and perform various operations like querying records via SOQL,managing metadata, CRUD operations and creating apex class without requiring custom integration for each use case.

Instead of hardcoding multiple API endpoints for each use case we can just use one MCP based interface.

Core Components of MCP

1. MCP Client

This is the AI interface Examples:

• VS Code Copilot

• Salesforce Agentforce

• Chat-based AI apps

👉 It decides what to do, but doesn’t directly touch data.

2. MCP Server

This is the bridge to real systems

It:

• Exposes tools (APIs, logic, queries)

• Executes actions securely

• Returns clean, structured responses

👉 Example tools:

• “Get Account details”

• “Create Case”

• “Fetch Opportunity pipeline”

3. Tools

Tools are actions AI is allowed to perform.

👉 Each tool defines:

• What it does

• Required inputs

• Output format

Example:

Tool: getAccountDetails
Input: accountId
Output: Name, Industry, Revenue

How MCP Works (Step-by-Step)

1. User asks a question “Show me account revenue”

2. AI (MCP Client) understands intent

3. AI selects the right tool

4. MCP Server runs real logic (Salesforce query, API call, Apex, etc.)

5. Structured data is returned

6. AI converts it into a human-friendly response

What You’re Building (Big Picture)

Flow:

VS Code (MCP Client)
   ↓
MCP Server (Node)
   ↓
Salesforce APIs (REST / Tooling / Metadata)

VS Code sends structured MCP requests → MCP server translates them → Salesforce responds → MCP sends results back to VS Code.

🛠 Prerequisites For Installing Salesforce Official Mcp Server

• Visual Studio Code installed.

• Node.js & npm installed.(Please prefer v22.21.1 (LTS) over newer versions due to some bugs related to mcp.)

• Salesforce CLI (sf or sfdx) installed and authenticated against at least one org (Dev Hub, scratch org, sandbox etc.).

• Optionally: GitHub Copilot Chat extension installed.

• Open a Salesforce DX project in VS Code (sfdx-project.json)

Step 1: Create MCP Configuration

In your project root, create a folder named .vscode (if it doesn’t exist). Inside that, create a file called:

.vscode/mcp.json

Then paste this configuration:

{
  "servers": {
    "Salesforce DX": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@salesforce/mcp@latest",
        "--orgs",
        "DEFAULT_TARGET_ORG",       
        "--toolsets",
        "all"
      ]
    }
  }
}

Step 2: Start the MCP Server in VS Code

• Open VS Code.

• Press Ctrl+Shift+P (Windows/Linux) or Cmd+Shift+P (macOS) → type:

• MCP: List Servers

• Select your configured server (e.g., Salesforce DX).

• Click Start Server.

• You should see logs in the Output → MCP panel showing the server is listening.

Step 3: Use an AI Client (Copilot / Agent Mode / Extension)

To actually talk to Salesforce you need an AI agent that uses the MCP server:

With Agentforce Chat

• Open Agentforce Chat window

• Switch to Act mode (not Plan).

• Now type natural language prompts, such as:

  “Show 5 most recent Account records.”

  “Show all open cases.”

  “Show recently closed cases.”

The agent figures out which MCP tool to call and gives you the result.

Also explore below link:

How to install the Custom MCP Server Part 2

Related Articles

• Build a UI which shows Agentforce Prompt Box and execute the command locally
• Creating an Apex Rest API and using it as a custom tool in Custom MCP Server
• How to add a Global Callable Class which Implements Callable and Register it as

Ready to implement? Learn more about our Agentforce Implementation Services.