Data 360 MCP Server

A Model Context Protocol server that exposes the Salesforce Data 360 Connect API to MCP-compatible clients such as Claude Code, Cursor, and any other host that speaks MCP over STDIO.

Built on Spring AI.

Developer Preview

The Data 360 MCP server is provided as a developer preview running in a single user/org context per process. We strongly encourage use only via STDIO instead of making this server network-accessible.

The generally available version of this MCP server is slated to be provided as a hosted and managed Salesforce Platform MCP server, alongside other existing Product Integration MCP servers.

During the developer preview please share feature requests, issues, success stories, or other feedback via github issues on this repo, the R&D team will be actively reviewing these on a regular basis. We're especially interested in feedback about how effective the fascade tools approach works across use cases and clients.

How it works

Data 360 exposes ~190 distinct REST operations. Rather than registering one MCP tool per endpoint (which quickly overwhelms an LLM's tool list and context window), this server consolidates everything behind three facade tools:

Tool	Purpose
search	Discover tools by intent, keyword, or family
payload_examples	Fetch working JSON payloads for complex tools
execute	Run any underlying tool by name with parameters

Typical workflow: search → payload_examples → execute.

Under the hood, 187 operations are organized into 21 tool families:

Family	Tools	Description
DLO	5	Data Lake Object CRUD
DMO	5	Data Model Object CRUD
Data Streams	9	Ingestion pipelines — generic, SFDC CRM, Snowflake, AWS S3
Mappings	7	DLO-to-DMO field mappings
Data Transforms	9	SQL transforms — CRUD, run, validate, schedule
Identity Resolution	8	IR rulesets — create, run, publish
Calculated Insights	10	CI CRUD, run, validate, query
Segment	6	Segment CRUD + publish
Dataspace	8	Dataspace + member management
Connection	11	Connector metadata, connection CRUD, Snowflake connections
Query	16	SQL, profile, datagraph, and metadata queries
Activation	10	Activation + target CRUD
SDM	38	Semantic Data Models — dimensions, measures, relationships, queries
Retriever	10	RAG retriever + configuration CRUD
DataKit	9	Bundle deploy / undeploy
Data Action	8	Action + target CRUD
Search Index	7	Vector / hybrid search index lifecycle
Smart Tools	4	AI-assisted mapping suggestions, event-date recommendations
Standard Mappings	2	Pre-configured mapping bundles
Eventing	2	Single + batch event publish
GDPR	3	Data subject read / delete requests

Quick start (one command)

The installer checks for Java 17+ and Maven (installs them if missing), clones the repo, builds the server JAR, prompts for your Salesforce credentials, and configures your MCP client (Claude Desktop, Claude Code, or Cursor):

macOS / Linux

curl -fsSL https://raw.githubusercontent.com/forcedotcom/d360-mcp-server/refs/heads/main/install.sh | bash

That's it. The script handles everything — including installing Java, Maven, and Git if you don't have them.

If you already cloned the repo and want to build the current checkout, run:

./install.sh

Windows (PowerShell)

Install Python 3.9+ first (the installer uses only the Python standard library). Then:

irm https://raw.githubusercontent.com/forcedotcom/d360-mcp-server/refs/heads/main/install.py -OutFile install.py
python install.py

The Python installer auto-installs Git, Java

... View full README on GitHub