MCP Force

You expose Salesforce APIs as tools for AI agents with MCP Force. This server runs locally or in your environment, connects to your Salesforce org, and makes REST, Bulk v2, GraphQL, Tooling, and Auth APIs accessible to agents like VS Code or Claude Desktop so you can query data and manage records via natural language.

How to use

Start by configuring your MCP client to point at the MCP Force server you run locally or in your environment. You will connect to Salesforce through the exposed tools, run queries, and manipulate records using natural language prompts in your editor or Claude Desktop.

Typical usage patterns include: 1) Connect to your Salesforce org with the MCP client. 2) Ask the agent to run a SOQL query, aggregate data, or manipulate records. 3) Review results returned by the agent with a clear breakdown and any relevant details. These steps work with VS Code in Agent mode or Claude Desktop when the MCP server is configured and running.

Key capabilities you can leverage include querying Salesforce data, executing SOQL queries, and performing standard CRM actions through natural language prompts. The MCP server provides access to REST, Bulk v2, GraphQL, Tooling, and Auth APIs depending on your configuration and selections.

How to install

Prerequisites you need before running MCP Force: Node.js 22+ and a Salesforce org you can connect to. You also need an access token for Salesforce and a target port for the MCP server.

1. Create a configuration file and environment setup. Place credentials in a .env file in your project directory with these variables:

SALESFORCE_ENDPOINT='https://mydomain-dev-ed.developer.my.salesforce.com'

# Run this command to login and obtain a token
auth:Salesforce token setup will be used by MCP Force
# Set your access token
SALESFORCE_ACCESS_TOKEN='<your_access_token_here>'

# Set your port (default shown as 3021 in examples)
PORT=3021

2. Start the MCP Force server using the MCP CLI. You can run with the default full API set or limit the APIs to reduce loaded tools.

npx @rapidocloud/mcp-force@1.0.0-dev.5 run-server

Alternative detailed startup that locks to specific APIs and uses a global installation step for MCP tools (shown as a complete command in setup examples):

npm install -g @rapidocloud/mcptools@1.0.0-dev.9 1>&2 && npx @rapidocloud/mcp-force@1.0.0-dev.5 run-server --selectedAPIs 'rest, tooling, metadata' --barredAPIs 'auth, bulk, graphql'

Additional configuration and usage notes

Configuration blocks show how to wire the MCP server into your development environment. The server can be integrated via a JSON-based MCP client config or within your editor’s MCP setup. Example configurations rely on the command and environment variables shown above.

Troubleshooting

If you encounter connection issues, verify that your Salesforce credentials are correct in the environment variables and that the access token is valid. Check that the MCP server is reachable from your client, and review logs for any authentication or host-related errors. Ensure you are using Node.js 22+ and that the port you configured (default 3021) is not blocked by a firewall.

Security and notes

This setup is experimental and intended for sandbox, developer, or scratch orgs. Do not use in production. Keep your access tokens secure and rotate them as needed. Monitor for updates to MCP tools and follow best practices for OAuth token handling.

Environment and tooling details

Environment variables you provide are the primary means to connect to Salesforce. You will typically include SALESFORCE_ENDPOINT, SALESFORCE_ACCESS_TOKEN, and PORT in your environment or through client configuration.

Available tools

list-tools

Lists all Salesforce API endpoints currently known to MCP Force

run-server

Launches the MCP Force server with selected APIs and barred APIs to expose Salesforce as tools for agents