Salesforce CLI MCP Server

Salesforce CLI MCP Server

Exposes Salesforce CLI functionality to LLM tools like Claude Desktop, allowing AI agents to execute Salesforce commands, manage orgs, deploy code, and query data through natural language.

Category
Visit Server

Tools

sf_cache_clear

sf_cache_refresh

sf_detect_project_directory

sf_set_project_directory

sf_list_roots

README

Salesforce CLI MCP Server

Model Context Protocol (MCP) server for providing Salesforce CLI functionality to LLM tools like Claude Desktop.

Overview

This MCP server wraps the Salesforce CLI (sf) command-line tool and exposes its commands as MCP tools and resources, allowing LLM-powered agents to:

  • View help information about Salesforce CLI topics and commands
  • Execute Salesforce CLI commands with appropriate parameters
  • Leverage Salesforce CLI capabilities in AI workflows

Requirements

  • Node.js 18+ and npm
  • Salesforce CLI (sf) installed and configured
  • Your Salesforce org credentials configured in the CLI

Installation

# Clone the repository
git clone <repository-url>
cd sfMcp

# Install dependencies
npm install

Usage

Starting the server

# Basic usage
npm start

# With project roots
npm start /path/to/project1 /path/to/project2
# or using the convenience script
npm run with-roots /path/to/project1 /path/to/project2

# As an npx package with roots
npx -y codefriar/sf-mcp /path/to/project1 /path/to/project2

The MCP server uses stdio transport, which can be used with MCP clients like the MCP Inspector or Claude Desktop.

Configuring in Claude Desktop

To configure this MCP in Claude Desktop's .claude.json configuration:

{
  "tools": {
    "salesforce": {
      "command": "/path/to/node",
      "args": [
        "/path/to/sf-mcp/build/index.js",
        "/path/to/project1",
        "/path/to/project2"
      ]
    }
  }
}

Using the npm package directly:

{
  "tools": {
    "salesforce": {
      "command": "/path/to/npx", 
      "args": [
        "-y",
        "codefriar/sf-mcp",
        "/path/to/project1",
        "/path/to/project2"
      ]
    }
  }
}

Development

# Watch mode (recompiles on file changes)
npm run dev

# In another terminal
npm start [optional project roots...]

Available Tools and Resources

This MCP server provides Salesforce CLI commands as MCP tools. It automatically discovers and registers all available commands from the Salesforce CLI, and also specifically implements the most commonly used commands.

Core Tools

  • sf_version - Get the Salesforce CLI version information
  • sf_help - Get help information for Salesforce CLI commands
  • sf_cache_clear - Clear the command discovery cache
  • sf_cache_refresh - Refresh the command discovery cache

Project Directory Management (Roots)

For commands that require a Salesforce project context (like deployments), you must specify the project directory. The MCP supports multiple project directories (roots) similar to the filesystem MCP.

Configuration Methods

Method 1: Via Command Line Arguments

# Start the MCP with project roots
npm start /path/to/project1 /path/to/project2
# or
npx -y codefriar/sf-mcp /path/to/project1 /path/to/project2

When configured this way, the roots will be automatically named root1, root2, etc., with the first one set as default.

Method 2: Using MCP Tools

  • sf_set_project_directory - Set a Salesforce project directory to use for commands
    • Parameters:
      • directory - Path to a directory containing a sfdx-project.json file
      • name - (Optional) Name for this project root
      • description - (Optional) Description for this project root
      • isDefault - (Optional) Set this root as the default for command execution
  • sf_list_roots - List all configured project roots
  • sf_detect_project_directory - Attempt to detect project directory from user messages

Example usage:

# Set project directory with a name
sf_set_project_directory --directory=/path/to/your/sfdx/project --name=project1 --isDefault=true

# List all configured roots
sf_list_roots

# Or include in your message:
"Please deploy the apex code from the project in /path/to/your/sfdx/project to my scratch org"

Method 3: Claude Desktop Configuration Configure project roots in .claude.json as described below.

Using Project Roots

You can execute commands in specific project roots:

# Using resource URI
sf://roots/project1/commands/project deploy start --sourcedir=force-app

# Using rootName parameter
sf_project_deploy_start --sourcedir=force-app --rootName=project1

Project directory must be specified for commands such as deployments, source retrieval, and other project-specific operations. If multiple roots are configured, the default root will be used unless otherwise specified.

Key Implemented Tools

The following commands are specifically implemented and guaranteed to work:

Organization Management

  • sf_org_list - List Salesforce orgs
    • Parameters: json, verbose
  • sf_auth_list_orgs - List authenticated Salesforce orgs
    • Parameters: json, verbose
  • sf_org_display - Display details about an org
    • Parameters: targetusername, json
  • sf_org_open - Open an org in the browser
    • Parameters: targetusername, path, urlonly

Apex Code

  • sf_apex_run - Run anonymous Apex code
    • Parameters: targetusername, file, apexcode, json
  • sf_apex_test_run - Run Apex tests
    • Parameters: targetusername, testnames, suitenames, classnames, json

Data Management

  • sf_data_query - Execute a SOQL query
    • Parameters: targetusername, query, json
  • sf_schema_list_objects - List sObjects in the org
    • Parameters: targetusername, json
  • sf_schema_describe - Describe a Salesforce object
    • Parameters: targetusername, sobject, json

Deployment

  • sf_project_deploy_start - Deploy the source to an org
    • Parameters: targetusername, sourcedir, json, wait

Dynamically Discovered Tools

The server discovers all available Salesforce CLI commands and registers them as tools with format: sf_<topic>_<command>.

For example:

  • sf_apex_run - Run anonymous Apex code
  • sf_data_query - Execute a SOQL query

For nested topic commands, the tool name includes the full path with underscores:

  • sf_apex_log_get - Get apex logs
  • sf_org_login_web - Login to an org using web flow

The server also creates simplified aliases for common nested commands where possible:

  • sf_get as an alias for sf_apex_log_get
  • sf_web as an alias for sf_org_login_web

The available commands vary depending on the installed Salesforce CLI plugins.

Note: Command discovery is cached to improve startup performance. If you install new SF CLI plugins, use the sf_cache_refresh tool to update the cache, then restart the server.

Resources

The following resources provide documentation about Salesforce CLI:

  • sf://help - Main CLI documentation
  • sf://topics/{topic}/help - Topic help documentation
  • sf://commands/{command}/help - Command help documentation
  • sf://topics/{topic}/commands/{command}/help - Topic-command help documentation
  • sf://version - Version information
  • sf://roots - List all configured project roots
  • sf://roots/{root}/commands/{command} - Execute a command in a specific project root

How It Works

  1. At startup, the server checks for a cached list of commands (stored in ~/.sf-mcp/command-cache.json)
  2. If a valid cache exists, it's used to register commands; otherwise, commands are discovered dynamically
  3. During discovery, the server queries sf commands --json to get a complete list of available commands
  4. Command metadata (including parameters and descriptions) is extracted directly from the JSON output
  5. All commands are registered as MCP tools with appropriate parameter schemas
  6. Resources are registered for help documentation
  7. When a tool is called, the corresponding Salesforce CLI command is executed

Project Roots Management

For commands that require a Salesforce project context:

  1. The server checks if any project roots have been configured via sf_set_project_directory
  2. If multiple roots are configured, it uses the default root unless a specific root is specified
  3. If no roots are set, the server will prompt the user to specify a project directory
  4. Commands are executed within the appropriate project directory, ensuring proper context
  5. The user can add or switch between multiple project roots as needed

Project-specific commands (like deployments, retrievals, etc.) will automatically run in the appropriate project directory. For commands that don't require a project context, the working directory doesn't matter.

You can execute commands in specific project roots by:

  • Using the resource URI: sf://roots/{rootName}/commands/{command}
  • Providing a rootName parameter to command tools (internal implementation details)
  • Setting a specific root as the default with sf_set_project_directory --isDefault=true

Command Caching

To improve startup performance, the MCP server caches discovered commands:

  • The cache is stored in ~/.sf-mcp/command-cache.json
  • It includes all topics, commands, parameters, and descriptions
  • The cache has a validation timestamp and SF CLI version check
  • By default, the cache expires after 7 days
  • When you install new Salesforce CLI plugins, use sf_cache_refresh to update the cache

Troubleshooting Cache Issues

The first run of the server performs a full command discovery which can take some time. If you encounter any issues with missing commands or cache problems:

  1. Stop the MCP server (if running)
  2. Manually delete the cache file: rm ~/.sf-mcp/command-cache.json
  3. Start the server again: npm start

This will force a complete rediscovery of all commands using the official CLI metadata.

If specific commands are still missing, or you've installed new SF CLI plugins:

  1. Use the sf_cache_refresh tool from Claude Desktop
  2. Stop and restart the MCP server

Handling Nested Topics

The Salesforce CLI has a hierarchical command structure that can be several levels deep. This MCP server handles these nested commands by:

  • Converting colon-separated paths to underscore format (apex:log:getsf_apex_log_get)
  • Providing aliases for common deep commands when possible (sf_get for sf_apex_log_get)
  • Preserving the full command hierarchy in the tool names
  • Using the official command structure from sf commands --json

Nested topic commands are registered twice when possible—once with the full hierarchy name and once with a simplified alias, making them easier to discover and use.

License

ISC

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured