Salesforce-Hosted-Custom-Mcp-Server
Enables interaction with Salesforce data and services via custom MCP tools, including account analytics, opportunity queries, case creation, and AI agent invocation.
README
Salesforce Hosted Custom MCP Server
A reference SFDX project demonstrating every supported pattern for building a Salesforce-hosted custom MCP (Model Context Protocol) server. Each McpServerDefinition shows a different way to expose Salesforce functionality — Apex classes, Flows, Named Queries, and Prompt Templates — as tools that any MCP-compatible client (Claude Desktop, OpenAI, Cursor, etc.) can discover and call.
Blog series: Salesforce Diaries by Sanket
What is a Salesforce Hosted MCP Server?
Salesforce lets you publish a standard MCP endpoint directly from your org — no middleware, no Lambda, no Node server. You define what tools are exposed through McpServerDefinition metadata, and Salesforce handles the authentication, schema generation, and HTTP transport. An external AI agent connects to https://<your-org>.my.salesforce.com/services/mcp/v1/<serverName> using OAuth and gets a live, permissioned connection to your data.
Project Structure
force-app/main/default/
├── mcpServerDefinitions/ # 14 MCP server definitions (one per pattern)
├── externalServiceRegistrations/ # OpenAPI registrations for each Apex endpoint
├── flows/
│ └── Get_Account_By_Name.flow-meta.xml
├── classes/ # 7 Apex classes backing the MCP tools
└── genAiPromptTemplates/ # 4 Prompt Builder templates exposed as MCP prompts
sfdx-project.json
MCP Server Definitions
Each definition registers one MCP server in your org and declares which tools and prompts it exposes.
1. FlowAsMcpService — Flow as MCP Tool
Exposes the Get_Account_By_Name AutoLaunchedFlow as a tool.
Pattern: fa:flow-<FlowApiName> (Flow via API Catalog)
Tool → Get_Account_By_Name
2. FlowMCP — Flow MCP (variant)
A second server exposing the same Get_Account_By_Name flow, demonstrating that multiple servers can surface the same underlying action.
Pattern: fa:flow-<FlowApiName>
3. InvocableApex — Invocable Apex + Prompt Templates
Combines an @InvocableMethod tool with three Prompt Builder prompts in one server.
Pattern: aa:apex-<ClassName> + promptTemplateName
Tool → CreateCaseAction
Prompt → Draft_IT_Troubleshooting
Prompt → Get_Event_Info
Prompt → Case_Research_from_Web
4. AuraEnabledApexMCP — @AuraEnabled Method
Exposes the getHighValueOpportunities method from OpportunityMcpService (annotated @AuraEnabled).
Pattern: ae:<ClassName>
Tool → getHighValueOpportunities (OpportunityMcpService)
5. AuraEnbaledClassMcp — @AuraEnabled (alternate class)
Same tool pattern as above but backed by OpportunityMcpServiceAuraEnbaled.
Pattern: ae:<ClassName>
Tool → getHighValueOpportunities (OpportunityMcpServiceAuraEnbaled)
6. PublicInvocableApexMCP — @AuraEnabled for Account Analytics
Exposes AccountAnalytics.invoke — a multi-mode analytics tool that runs aggregations over Account records.
Pattern: ae:<ClassName>
Tool → invoke (AccountAnalytics)
7. AIAgentMcpServer — Talk to an Agentforce Agent
Lets an MCP client send a natural-language message to an Agentforce agent and get its reply. Supports multi-turn sessions via sessionId.
Pattern: ae:<ClassName>
Tool → askAgent (InvokeAgentAction)
8. CaseCreationInvocableMcpService — Create Cases via Invocable
Exposes CreateCaseAction as an MCP tool so external agents can open support cases directly.
Pattern: aa:apex-<ClassName>
Tool → CreateCaseAction
9. NamedQueryMcpService — Named Query
Exposes a Named Query (GetOpportunitiesByAmount) defined in the org as a zero-code MCP tool.
Pattern: nq:<QueryApiName>
Tool → GetOpportunitiesByAmount
10. PromptTemplateMcpServer — Named Query + Prompt Template
Mixes a Named Query tool with a Prompt Builder prompt in one server.
Pattern: nq:<QueryApiName> + promptTemplateName
Tool → GetOpportunitiesByAmount
Prompt → Research_on_Case_subject
11. RestResourceApexMCP — @RestResource Apex
Exposes MyCustomAPI — a @RestResource-annotated class — as an MCP tool.
Pattern: ar:<ClassName>
Tool → doGet (MyCustomAPI)
12. PromptMcpServer — Prompt-Only Server
A server that exposes only a Prompt Builder template, with no tool.
Pattern: promptTemplateName
Prompt → Case_Research_from_Web
13. AgentforceMCP — Agentforce Agent Shell
Skeleton server demonstrating how to declare an Agentforce-connected MCP server.
14. RestAPIMCP — REST API Shell
Skeleton server for REST-based MCP tool patterns.
Apex Classes
| Class | Annotation | What it does |
|---|---|---|
OpportunityMcpService |
@AuraEnabled |
Returns Opportunities with Amount > threshold (default $50k), ordered largest first. Enforces sharing and FLS via WITH USER_MODE. |
OpportunityMcpServiceAuraEnbaled |
@AuraEnabled |
Variant of the above, demonstrating the same pattern in a separate class. |
AccountAnalytics |
@InvocableMethod |
Runs one of four aggregation modes over Account: top_revenue, by_industry, recent, or health_summary. Returns a human-readable summary string. |
CreateCaseAction |
@InvocableMethod |
Creates a Salesforce Case from a subject line. Bulk-safe; re-queries to return the auto-assigned Case Number. |
InvokeAgentAction |
@InvocableMethod |
Sends a natural-language message to any active Agentforce agent via generateAiAgentResponse. Returns the agent's reply and a sessionId for multi-turn conversations. Default agent: Case_CRUD_Agent. |
MyCustomAPI |
@RestResource |
GET /services/apexrest/MyCustomAPI/ — returns Opportunities above a configurable minAmount threshold. Used by RestResourceApexMCP. |
PromptTemplateMcpService |
@AuraEnabled |
Runs a named Prompt Builder template against a record using ConnectApi.EinsteinLLM.generateMessagesForPromptTemplate. Returns the generated text. |
Flow
Get_Account_By_Name
Type: AutoLaunchedFlow
Input: AccountName (String)
Output: AccountDetails (Account SObject)
Queries the first Account whose Name equals the input string and returns the full record. Used by FlowAsMcpService and FlowMCP.
Prompt Templates (genAiPromptTemplates)
These Prompt Builder templates are registered on MCP servers as prompts — the MCP equivalent of a reusable, parameterised instruction that a client can invoke by name.
| Template | Referenced by |
|---|---|
Draft_IT_Troubleshooting |
InvocableApex |
Get_Event_Info |
InvocableApex |
Case_Research_from_Web |
InvocableApex, PromptMcpServer |
Research_on_Case_subject |
PromptTemplateMcpServer |
External Service Registrations
OpenAPI-based registrations that describe the shape of each Apex endpoint so Salesforce can generate the tool schemas for the MCP catalog.
| Registration | Apex class |
|---|---|
AccountAnalytics |
AccountAnalytics |
GetCaseByCaseNumber |
Case lookup endpoint |
GetOpportunitiesByAmount |
Named Query |
InvokeAgentAction |
InvokeAgentAction |
MyCustomAPI |
MyCustomAPI |
OpportunityMcpService |
OpportunityMcpService |
OpportunityMcpServiceAuraEnbaled |
OpportunityMcpServiceAuraEnbaled |
PromptTemplateMcpService |
PromptTemplateMcpService |
Deploy
Prerequisites
- Salesforce CLI (
sf) v2+ - API version 66.0 org (Summer '25+)
- "Salesforce Hosted MCP Servers" feature enabled in your org
Deploy all metadata
sf project deploy start --source-dir force-app
Deploy a specific server only
sf project deploy start --metadata McpServerDefinition:FlowAsMcpService
Connect a client (example: Claude Desktop)
Add this to your claude_desktop_config.json:
{
"mcpServers": {
"salesforce-flow": {
"url": "https://<your-org>.my.salesforce.com/services/mcp/v1/FlowAsMcpService",
"transport": "http",
"headers": {
"Authorization": "Bearer <access_token>"
}
}
}
}
References
Recommended Servers
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.
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.
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.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
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.
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.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
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.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.