MCP Force
Expose Salesforce APIs as tools for AI agents, enabling natural language interaction with Salesforce organizations.
MCP Force : an MCP server to expose Salesforce APIs as tools for AI Agents
This program provides an MCP (Model Context Protocol) server that exposes Salesforce APIs as tools for use in AI agents. It enables AI assistants like Claude Desktop and VS Code to interact with your Salesforce organization through natural language.
Currently, the following APIs are covered :
- Salesforce Platform APIs : REST API, Bulk v2 API, GraphQL API, Tooling API, Auth API
EXPERIMENTAL
This package is experimental.
Please use it ONLY in a sandbox, a developer org or a scratch org. NEVER use this in production : you have been warned.
The APIs provided by MCP Force are defined in the @rapidocloud/mcp-tools package on NPMjs.org.
All Salesforce APIs will be connected, depending on user requests : open an issue to ask for more.
Finally, user authentication is manual, for the time being : set an OAuth2 access token in the .env file. This will be improved very soon.
Features
- Connect to Salesforce organizations
- Query and manipulate Salesforce data
- Execute SOQL queries
- Manage Salesforce records
- Integration with VS Code and Claude Desktop
Setup
1. Environment Configuration
Create a .env file in your project directory with your Salesforce credentials:
SALESFORCE_ENDPOINT='https://mydomain-dev-ed.developer.my.salesforce.com'
# run this command :
# sf org login web --instance-url ${SALESFORCE_ENDPOINT} --json
SALESFORCE_ACCESS_TOKEN='<your_access_token_here>'
# Set your port : 3001 by default
PORT=3021
3. Command Line Usage
Display the help menu with available commands and options
npx @rapidocloud/[email protected] --help
List all the Salesforce API endpoints currently known to MCP Force
npx @rapidocloud/[email protected] list-tools
Run a specific version of MCP Force
npx @rapidocloud/[email protected]
Run the MCP server in default mode with all APIs selected
npx @rapidocloud/mcp-force run-server
Run the MCP server only with specified APIs to reduce the number of tools loaded
npx @rapidocloud/mcp-force run-server --selectedAPIs 'rest,tooling'
Run the MCP server with all APIs except for specified ones
npx @rapidocloud/mcp-force run-server --barredAPIs 'bulk'
Run the MCP server with all API endpoints except for specified tool names
npx @rapidocloud/mcp-force run-server --barredToolNames 'run_tests_sync, tooling_run_tests_async'
Run the MCP server with a specific version of MCP Tools, selected APIs and barred APIs
npm install -g @rapidocloud/[email protected] 1>&2 && npx @rapidocloud/[email protected] run-server --selectedAPIs 'rest, tooling, metadata' --barredAPIs 'auth, bulk, graphql'
To use this in your MCP server configuration :
{
"servers": {
"mcpforce": {
"command": "bash",
"args": ["-c", "npm install -g @rapidocloud/[email protected] 1>&2 && npx @rapidocloud/[email protected] run-server --selectedAPIs 'rest, tooling, metadata' --barredAPIs 'auth, bulk, graphql'"],
"env": {
"SALESFORCE_ENDPOINT": "https://mydomain-dev-ed.developer.my.salesforce.com",
"SALESFORCE_ACCESS_TOKEN": "<your_access_token_here>",
"PORT": "3021"
}
}
}
}
4. VS Code Configuration
Create or update .vscode/mcp.json to configure the MCP server:
{
"servers": {
"mcp-force": {
"command": "npx",
"args": ["-y", "@rapidocloud/mcp-force"],
"env": {
"SALESFORCE_ENDPOINT": "https://mydomain-dev-ed.developer.my.salesforce.com",
"SALESFORCE_ACCESS_TOKEN": "<your_access_token_here>",
"PORT": "3021"
}
}
}
}
5. Claude Desktop Configuration
Add the MCP server to your Claude Desktop configuration file:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"mcp-force": {
"command": "npx",
"args": ["-y", "@rapidocloud/mcp-force"],
"env": {
"SALESFORCE_ENDPOINT": "https://mydomain-dev-ed.developer.my.salesforce.com",
"SALESFORCE_ACCESS_TOKEN": "<your_access_token_here>",
"PORT": "3021"
}
}
}
}
Installation
No installation required! The MCP server will be automatically downloaded and run using npx when configured in your MCP client.
For development or testing, you can also run it directly:
npx @rapidocloud/mcp-force list-tools
npx @rapidocloud/mcp-force run-server
Usage Examples
VS Code Agent Mode
In VS Code with the MCP extension and Agent mode enabled:
- Open the Command Palette (
Cmd/Ctrl + Shift + P) - Select "MCP: Start Agent Session"
- Ask the agent:
Sum the amount of all open Salesforce opportunities
The agent will:
- Connect to your Salesforce org
- Query for open opportunities using SOQL
- Calculate the total amount
- Provide you with the sum and breakdown
Claude Desktop Agent Mode
In Claude Desktop with the MCP server configured:
- Start a new conversation
- Ask Claude:
Sum the total amount of all open Salesforce opportunities?
Connect to my Salesforce org and run a query to get this information.
Claude will:
- Use the MCP tools to connect to Salesforce
- Execute a SOQL query like:
SELECT Amount FROM Opportunity WHERE IsClosed = false AND Amount != null - Sum the amounts and provide a detailed report
Example response:
I've connected to your Salesforce org and found the following open opportunities:
Total Open Opportunity Amount: $2,450,000
Breakdown:
- 15 opportunities with amounts ranging from $25,000 to $500,000
- Average deal size: $163,333
- Largest opportunity: $500,000
- Smallest opportunity: $25,000
Troubleshooting
- Ensure your Salesforce credentials are correct in the environment variables
- Verify that your security token is up to date
- Check that the MCP server is properly configured in your client application
- Review the console logs for connection errors
- Make sure you have Node.js 22+ installed for npx to work properly
Requirements
- Node.js 22+
- Valid Salesforce credentials
- VS Code with MCP extension (for VS Code usage)
- Claude Desktop (for Claude Desktop usage)
Verwandte Server
notebooklm-mcp-secure
Security-hardened NotebookLM MCP with post-quantum encryption, GDPR/SOC2 compliance, and 14 security layers. Query Google's Gemini-grounded research from any MCP-compatible AI assistant.
EdgeOne Pages MCP
An MCP service for deploying HTML content to EdgeOne Pages and obtaining a publicly accessible URL.
Giant Swarm App Platform
A server for managing Giant Swarm App Platform deployments using Kubernetes credentials.
AWS CloudTrail
This AWS Labs Model Context Protocol (MCP) server for CloudTrail enables your AI agents to query AWS account activity for security investigations, compliance auditing, and operational troubleshooting.
Azure AHDS FHIR MCP Server
An MCP server implementation for interacting with Azure Health Data Services FHIR.
MCP Server for National Park Services Data
Retrieve data from the National Park Service (NPS) API.
AWS MCP Servers
A suite of MCP servers providing AI applications with access to AWS documentation, contextual guidance, and best practices.
ConvertAPI
High-Performance File Conversion MCP
ServiceNow MCP Server
An MCP server for interfacing with ServiceNow, enabling AI agents to access and manipulate data via a secure API.
Kaggle
Interact with the Kaggle API to access datasets, notebooks, and competitions.