Agentforce Architects - MCP Server for Salesforce Metadata

MCP Tools

The MCP Server provides a comprehensive set of tools for interacting with Salesforce metadata through natural language.

Connection Tools

connect_to_salesforce

Establishes a connection to your Salesforce organization using your credentials.

Key Features:

Supports username/password authentication
Handles security token automatically
Verifies connection with test query
Maintains session for subsequent operations

Technical Details:

Uses simple-salesforce library with multiple authentication methods. Attempts both separate token and password+token concatenation methods. Performs test query to confirm connection validity.

Sample Prompt:

Connect to my Salesforce org using my credentials

get_salesforce_connection_status

Checks if the current Salesforce connection is active and valid.

Key Features:

Verifies session token validity
Returns organization details
Identifies expired sessions
Provides connection diagnostics

Technical Details:

Executes a simple query against Organization object. Verifies session token is still valid. Identifies expired sessions before operations fail.

Sample Prompt:

Check if I'm still connected to Salesforce

Metadata Tools

describe_metadata_types

Lists all available metadata types in your Salesforce organization.

Key Features:

Retrieves complete metadata type catalog
Includes XML names and directory information
Shows metadata type descriptions
Supports filtering by category

Technical Details:

Uses REST Metadata API's describe endpoint. Provides fallback mechanism with predefined common types if API call fails. Filters and formats metadata type information for clarity.

Sample Prompt:

What metadata types are available in my Salesforce org?

list_metadata

Retrieves metadata components of a specific type from your Salesforce organization.

Key Features:

Lists components by metadata type
Includes component names and details
Supports folder-based metadata
Handles large result sets

Technical Details:

Uses SOAP Metadata API's listMetadata operation. Parses complex SOAP responses into structured data. Handles folder-based metadata types properly.

Sample Prompt:

List all custom objects in my Salesforce org

describe_object

Provides detailed information about a specific Salesforce object's schema.

Key Features:

Shows all fields and their properties
Includes relationships and child objects
Displays field-level security information
Shows record type details

Technical Details:

Uses REST API's describe endpoint for object schema. Returns complete object structure with all fields and relationships. Caches results for frequently accessed objects.

Sample Prompt:

Describe the Account object schema

create_custom_field

Creates a new custom field on a Salesforce object.

Key Features:

Supports all field types (text, number, date, etc.)
Configures field properties and validation
Sets field-level security
Handles dependent picklists

Technical Details:

Uses Tooling API for faster field creation. Handles conversion between common names and Salesforce API types. Builds appropriate metadata structure based on field type. Ensures field names meet Salesforce naming conventions.

Sample Prompt:

Create a new text field called 'Customer Priority' on the Account object

create_custom_object

Creates a new custom object in your Salesforce organization.

Key Features:

Configures object properties and behavior
Sets up standard fields automatically
Configures object permissions
Supports record types and page layouts

Technical Details:

Uses Metadata API for reliable object creation. Constructs proper XML structure for object definition. Returns deployment ID for status monitoring. Enforces proper API naming rules.

Sample Prompt:

Create a new custom object called 'Project' with fields for name, start date, and status

create_data_model

Creates a complete data model with multiple objects and relationships.

Key Features:

Builds multiple related objects
Creates lookup and master-detail relationships
Configures field dependencies
Deploys as a single transaction

Technical Details:

Uses Metadata API's deploy operation for transactional deployment. Properly configures lookup and master-detail relationships. Provides detailed error information for failed deployments. Verifies model specification before deployment.

Sample Prompt:

Create a project management data model with Projects, Tasks, and Resources objects

SOQL Tools

generate_soql_query

Creates a SOQL query based on natural language description.

Key Features:

Translates natural language to SOQL
Handles complex query conditions
Supports relationship queries
Includes proper field selection

Technical Details:

Constructs SOQL strings with proper escaping. Can automatically select fields if not specified. Sanitizes inputs to prevent SOQL injection. Ensures generated queries follow SOQL syntax rules.

Sample Prompt:

Create a query to find all Accounts with more than $1M in revenue

execute_soql_query

Executes a SOQL query against your Salesforce organization.

Key Features:

Runs queries and returns results
Handles large result sets with pagination
Formats results for readability
Provides error diagnostics

Technical Details:

Uses REST API's query endpoint. Removes Salesforce metadata from results for cleaner output. Identifies when results are paginated. Provides clear error messages for invalid queries.

Sample Prompt:

Run a query to find all Opportunities closed this month

Test Data Tools

generate_test_data

Creates realistic test data records in your Salesforce organization.

Key Features:

Generates data for any object
Creates related records automatically
Supports custom field values
Handles bulk data creation

Technical Details:

Uses Bulk API for efficient data insertion. Determines required fields and their constraints through schema analysis. Creates appropriate values based on field types. Supports industry-specific or contextual data generation.

Sample Prompt:

Generate 50 test Account records with related Contacts

Search Documentation Tool (RAG Solution)

search_salesforce_documentation

Searches official Salesforce documentation to find relevant information and best practices.

Key Features:

Searches official Salesforce documentation
Returns relevant articles and guides
Provides direct links to documentation
Summarizes key information

Technical Details:

Uses Google Custom Search API with Salesforce-specific scope. Structures search results with titles, links, and snippets. Provides clear messages for API errors or rate limiting. Requires Google API key and Custom Search Engine ID.

Sample Prompt:

Find documentation about Salesforce formula fields

Utility Tools

get_org_limits

Retrieves current usage and limits for your Salesforce organization.

Key Features:

Shows API usage and limits
Displays storage utilization
Reports on governor limits
Helps diagnose potential capacity issues

Technical Details:

Uses the REST API's limits endpoint to retrieve current organization limits and usage. Simplifies complex limits response for readability. Lightweight operation with minimal API consumption.

Sample Prompt:

Check my Salesforce org's current API usage and limits

MCP vs. AgentForce Comparison

MCP Server Approach

Direct Integration: Connects Claude directly to Salesforce APIs without intermediaries, eliminating additional platform dependencies
Lightweight Development: Minimal infrastructure requirements with simple Python-based implementation that runs locally with Claude desktop client
Unified Development Experience: Add new functions in a single file with consistent code structure, compared to creating multiple Apex classes and agent configurations
Open Protocol: Based on the standardized Model Context Protocol, enabling greater interoperability and future-proofing your implementation
Developer-Friendly: Python-based implementation familiar to Salesforce developers with straightforward debugging and monitoring capabilities
Rapid Iteration: Make changes and deploy immediately without going through multiple configuration screens or waiting for platform updates

AgentForce Approach

Platform-Based: Requires AgentForce platform as intermediary with multiple configuration screens for each component
Multi-Step Configuration: Necessitates creation of separate agent topics, agent actions, and Apex classes that must be associated through configuration
Fragmented Development: Requires working across multiple development environments (Apex classes, Configuration UI, Agent Builder)
Deployment Complexity: Changes require updates to multiple components and potential redeployment across different parts of the system
User-Friendly: Designed with administrators in mind with GUI-based configuration options that require less coding knowledge
Enterprise Support: Includes Salesforce's standard support channels, documentation, and regular platform updates

Key Differentiators of the MCP Approach

The MCP Server approach provides greater flexibility, control, and development efficiency for technical teams by:

Streamlining the Development Process: Instead of creating separate Apex classes for each agent action and then associating them with an AgentForce agent through multiple configuration steps, developers can simply add new functions in the same Python file with consistent syntax.
Reducing Implementation Complexity: The MCP implementation avoids the multi-step process required in AgentForce (creating agent topics, writing Apex classes, configuring agent actions, and associating those actions with the agent).
Enabling Faster Iteration: Direct control over the entire stack allows for rapid modifications and testing without navigating through multiple platform layers or waiting for changes to propagate.

Download MCP Server

Get the Python MCP Server to integrate with your Claude desktop client.

Download MCP Server

Setup Instructions

Download and extract the MCP Server ZIP file
Install required Python dependencies:
```
pip install -r requirements.txt
```

Create a .env file in the same directory as the server with the following credentials:

# Salesforce Credentials
SF_USERNAME=your_salesforce_username
SF_PASSWORD=your_salesforce_password
SF_SECURITY_TOKEN=your_salesforce_security_token

# Google API Credentials (for Search Documentation tool)
GOOGLE_API_KEY=your_google_api_key
GOOGLE_CSE_ID=your_google_custom_search_engine_id

Run the server locally:
```
python secure_mcp_server.py
```
Configure your Claude desktop client to use the MCP server:
1. Create a claude_desktop_config.json file with the following structure:
```
{
  "mcpServers": {
    "salesforce": {
      "command": "python3",
      "args": [
        "/path/to/your/downloaded/secure_mcp_server.py"
      ]
    }
  }
}
```
  Replace /path/to/your/downloaded/ with the actual path where you extracted the MCP server files.
2. Place this config file in the appropriate location for your Claude desktop client
3. Restart your Claude desktop client to connect to the MCP server

For detailed technical information, see the included technical_details.md file.

Intelligent Salesforce Metadata Orchestration via Claude AI & MCP

MCP Architecture

MCP Tools

Connection Tools

connect_to_salesforce

Key Features:

Technical Details:

Sample Prompt:

get_salesforce_connection_status

Key Features:

Technical Details:

Sample Prompt:

Metadata Tools

describe_metadata_types

Key Features:

Technical Details:

Sample Prompt:

list_metadata

Key Features:

Technical Details:

Sample Prompt:

describe_object

Key Features:

Technical Details:

Sample Prompt:

create_custom_field

Key Features:

Technical Details:

Sample Prompt:

create_custom_object

Key Features:

Technical Details:

Sample Prompt:

create_data_model

Key Features:

Technical Details:

Sample Prompt:

SOQL Tools

generate_soql_query

Key Features:

Technical Details:

Sample Prompt:

execute_soql_query

Key Features:

Technical Details:

Sample Prompt:

Test Data Tools

generate_test_data

Key Features:

Technical Details:

Sample Prompt:

Search Documentation Tool (RAG Solution)

search_salesforce_documentation

Key Features:

Technical Details:

Sample Prompt:

Utility Tools

get_org_limits

Key Features:

Technical Details:

Sample Prompt:

Security Considerations

Important

Recommended Usage

Key Security Practices

MCP vs. AgentForce Comparison

MCP Server Approach

AgentForce Approach

Key Differentiators of the MCP Approach

Download MCP Server

Setup Instructions

Try It Out

Example Prompts