Case Study · Salesforce Custom MCP

A unified AI layer for Salesforce, deployed in production.

WNC engineered a custom Model Context Protocol (MCP) server for a leading hospitality sales outsourcing enterprise – connecting Salesforce, Slack, Teams and agent frameworks so 110+ Remote Directors of Sales can query accounts, opportunities and activity in natural language, without leaving the tools where they actually work.

Industry

Hospitality

Region

USA · Canada · LATAM

Platforms

Salesforce · Slack · Teams · Agentforce

Practice

Custom MCP Server Engineering

3,500+

Hotels In Scope

110+

Remote Sales Users

7,000+

Hotels Served Daily

1

Unified AI Surface

01 · Client

The hospitality industry's leading remote hotel sales outsourcing company.

Founded in 2007 and repeatedly named to the Inc. 5000 list of fastest-growing US companies, the customer powers Remote Directors of Sales for over 3,500 hotels across the US, Canada and Latin America – spanning brands including Hilton, Marriott, IHG and Hyatt. Their highly distributed teams operate across multiple sales platforms, brand tools and CRM systems every single day.

02 · The Work

Bridging fragmented sales tools with a single, governed AI fabric.

The Challenge

110+ sellers. Dozens of systems. No unified intelligence layer.

  • Remote sales teams worked across Salesforce, brand hotel platforms, GDS systems, OTA tools and internal reporting – no AI layer could see across all of them.
  • Each Regional Director of Sales used different tools per hotel brand, making a unified AI assistant impossible with standard one-to-one integrations.
  • Sales intelligence and booking activity lived inside Salesforce but needed to surface inside Slack and Teams – where the team actually worked.
  • With 110+ sellers making daily calls across 7,000+ hotels, response speed and data accuracy were non-negotiable.

The Solution

A production-grade MCP server, engineered by WNC.

  • Deployed a Production MCP Server as the connector between Salesforce and Slack, Teams and other agent frameworks the team already used.
  • Remote Directors of Sales now query accounts, opportunities and activity in natural language – without ever opening Salesforce.
  • A smart query builder and metadata caching deliver fast, accurate responses regardless of org complexity.
  • User-context execution preserves Salesforce sharing rules – each RDOS only sees the hotel accounts they're assigned to.
  • Agentforce actions exposed as MCP tools let AI agents trigger follow-up workflows and create tasks directly from Slack.

03 · Architecture

One protocol, any AI surface, governed by your Salesforce org.

MCP — the Model Context Protocol — is the emerging open standard for how AI agents connect to tools and data. WNC's MCP server treats Salesforce as the source of truth and exposes it as a set of typed, governed tools any MCP-aware client can call.

Slack

Native chat surface for sellers

Microsoft Teams

Field & ops collaboration

Agent Frameworks

Claude, Cursor, Codex, custom agents

Agentforce

Salesforce-native AI actions

WNC

Production MCP
Server

Smart query · Metadata cache · User-context exec · Agentforce tools

SOQL builder
Schema cache
Tool registry
Sharing rules
Audit log
Rate limits

Salesforce CRM

Accounts · Opportunities · Activities

Agentforce Actions

Workflows, tasks, follow-ups

Metadata API

Objects, fields, picklists

Reports & Analytics

Pipeline, booking pace, KPIs

04 · Engineering Depth

What makes a production-grade MCP server different.

[ 01 ]

Smart Query Builder

Translates natural-language intent into safe, optimized SOQL – handling joins, governor limits and field-level visibility without hand-written queries per use case.

[ 02 ]

Metadata Caching

Fingerprinted schema cache eliminates redundant Describe calls. Responses stay fast even as the org grows in objects, custom fields and validation rules.

[ 03 ]

User-Context Execution

Every tool call runs under the calling user's Salesforce identity. Record-level security, sharing rules and territory assignments are preserved end-to-end.

[ 04 ]

Agentforce as Tools

Existing Agentforce actions are exposed as typed MCP tools – letting any AI client trigger Salesforce-native workflows without rebuilding business logic.

[ 05 ]

Multi-Surface Delivery

One server, many clients: Slack, Teams, Claude, Cursor, Agentforce or any future MCP-compatible AI surface. Plug in once, expose everywhere.

[ 06 ]

Observability & Audit

Every tool invocation is logged with the user, prompt, query, latency and result – giving admins a defensible audit trail and a tuning surface for LLM usage.

05 · Headless 360

Built for the Headless 360 era of Salesforce.

Salesforce Headless 360 opens the platform up to AI agents and developers through APIs, hosted MCP servers, CLI tools and a Headless Experience Layer that separates what an agent does from how it renders. WNC plays in this ecosystem in two complementary ways.

// WNC × HEADLESS 360
ANGLE 01 · PARTNER

We build on top of Salesforce's Headless 360.

For customers who want to standardise on Salesforce-hosted MCP servers, WNC designs the surrounding experience layer – connecting Headless 360 tools into Slack, Teams, mobile and custom React surfaces, and wiring governance, telemetry and prompt design around it.

  • Adopt Salesforce-hosted MCP tools (60+ out of the box).
  • Compose them with custom Apex, Flows and Data Cloud assets.
  • Native rollout into Slack, Teams, mobile and ChatGPT / Claude / Gemini.
  • DevOps, observability and centralized LLM governance.

ANGLE 02 · CUSTOM

We build production MCP servers when Headless 360 alone isn't enough.

Some organisations need behaviour that goes beyond the standard Headless 360 tool set – multi-tenant data shaping, brand-specific business logic, deep caching strategies, or interaction patterns specific to their industry. That's where WNC's custom MCP server engineering steps in – as the hospitality case study shows.

  • Custom SOQL & query orchestration for complex orgs.
  • Bespoke tool design with strong typing and validation.
  • Edge caching tuned to your usage patterns and SLAs.
  • Compatible with the same MCP clients Headless 360 supports.

06 · Outcomes

From fragmented stacks to one conversational surface.

Unified AI

One AI layer across the tool stack.

Sellers no longer hop between Salesforce, brand portals and reporting tools. Salesforce becomes the source of truth, MCP becomes the surface.

Zero Context Switching

Salesforce data, inside Slack and Teams.

Remote Directors of Sales ask in natural language and receive accounts, opportunities and activity directly inside the tools where they already work.

Performance

Dramatically faster responses through caching.

Metadata and query caching cut redundant API traffic and keep latency low — critical for a sales team making thousands of decisions a day.

Future-Proof

One server, any future AI tool.

As new MCP-aware AI tools emerge, the customer plugs them straight into their existing Salesforce data and Agentforce actions – no rebuilds.

"WNC has been our trusted Salesforce partner for nearly 10 years — consistently professional, adaptable and problem-solving. Their flexible, versatile approach has made them an invaluable resource."

Customer Leadership · Hospitality sales outsourcing enterprise · USA