Decision 01
Operating job
Who needs to do what, in which system, with what evidence?
Salesforce MCP · Architecture · Governance · Adoption
Connect AI to Salesforce with the right MCP architecture
GAT assesses, configures, governs, and launches Salesforce MCP for Claude, ChatGPT, and other AI work surfaces—from Salesforce Hosted MCP and ApexGenius to a custom server. The use case chooses the path; security and operating ownership are designed before launch.
Decision 02
Trust boundary
Which identity, data, tools, writes, approvals, and environments are allowed?
Decision 03
Working path
Hosted MCP, ApexGenius, Agentforce, custom MCP—or a deliberate combination.
Product, native platform, or custom build
Four paths can all be right. They are not the same job.
GAT implements across the options and built one of them. We disclose that product interest, compare the practical boundaries, and recommend the smallest architecture that safely completes the workflow.
Salesforce Hosted MCP
Use the native Salesforce-managed path
A strong fit when supported Salesforce endpoints cover the required data and actions and the organization wants the connection governed inside its existing Salesforce footprint.
- Edition and entitlement validation
- External client and OAuth setup
- Permission-set and action scoping
ApexGenius
Use the packaged Salesforce connector
A strong fit when admins, developers, consultants, or operators need Salesforce records plus deeper schema, metadata, Apex, Flow, test, and deployment workflows from their AI client.
- Hosted product by GAT
- Claude, ChatGPT, and compatible MCP clients
- Product setup instead of custom server ownership
Custom MCP
Build around a specialized operating job
A strong fit when the workflow crosses multiple systems, needs purpose-built tools, or has network, identity, approval, tenancy, or user-experience requirements a packaged connector cannot satisfy.
- Custom tools and resource design
- Cross-system orchestration
- Evaluations, monitoring, and operations
Agentforce
Build an agent inside Salesforce
A strong fit when the goal is a customer- or employee-facing agent operating in Salesforce channels, rather than primarily connecting an external AI work surface to the org.
- Use-case and platform fit
- Topics, actions, testing, and guardrails
- Consumption and operating model
Implementation path
The connection is only one layer of the launch
Production value comes from the operating job, trust boundary, complete workflow, and adoption model around the connector.
01
Define the operating job
Choose the users, decisions, records, metadata, actions, environments, and business outcome before choosing a connector.
02
Design the trust boundary
Map identity, OAuth, profiles, permission sets, field access, approved tools, human review, auditability, and revocation.
03
Prove the complete loop
Connect a sandbox and test the real workflow end to end—including bad inputs, denied actions, timeouts, and recovery paths.
04
Launch and operate
Enable the right users, document safe patterns, measure adoption and quality, and maintain permissions and workflows as the org changes.
Work worth connecting
Give AI a real job inside Salesforce
The target is not a generic chat window. It is a permission-aware working loop that reduces context hunting, completes a bounded action, or makes a business decision easier to inspect.
Org discovery
Inspect objects, fields, relationships, automations, and dependencies before scoping a change.
Revenue & service analysis
Query current pipeline, customer, case, or operational data without moving the work into an exported spreadsheet.
Admin workflows
Document configuration, investigate failures, stage safe changes, and keep the operating context attached to the request.
Development workflows
Understand Apex and Flow behavior, prepare metadata, validate changes, run tests, and support reviewed deployment paths.
AI-enabled operations
Give Claude or ChatGPT approved tools for intake, routing, summarization, follow-up, reporting, or case assistance.
Cross-system agents
Combine Salesforce context with documents, support platforms, billing, internal APIs, or custom user experiences.
Cost model
Separate the AI, connector, Salesforce, and implementation bills
MCP pricing gets confusing when four different layers are presented as one number. GAT scopes the implementation and governance work; ApexGenius publishes its current connector subscription on the product site.
Questions
Salesforce MCP consulting FAQ
What does Salesforce MCP consulting include?
It can include option selection, current-state architecture, Salesforce and AI-client setup, identity and permission design, tool and action scoping, sandbox validation, workflow engineering, documentation, enablement, evaluations, monitoring, and ongoing operation.
Do we need a custom MCP server?
Not necessarily. GAT starts with the operating job and compares Salesforce Hosted MCP, ApexGenius, Agentforce, and a custom server. A packaged path is usually preferable when it safely covers the required users, systems, and actions.
How is this different from ApexGenius?
ApexGenius is the packaged Salesforce MCP product GAT built. MCP consulting is the architecture and implementation service: deciding the right path, configuring governance, redesigning workflows, connecting other systems, or engineering a custom solution when the product alone is not the whole answer.
Can GAT connect both Claude and ChatGPT?
Yes. The design can support Claude, ChatGPT, compatible developer tools, or a mixed environment. The work surface is selected by user and workflow; the Salesforce permission model and action boundary remain explicit either way.
Can MCP change Salesforce records or metadata?
That depends on the server, enabled tools, connected Salesforce user, and review model. GAT starts read-only where appropriate, scopes write and deployment capabilities deliberately, and validates consequential paths in a sandbox before production use.
What does Salesforce MCP cost?
There can be separate costs for the AI client, the MCP connector or platform, Salesforce services and usage, and implementation or ongoing governance. GAT publishes a practical cost framework and ApexGenius maintains its current product pricing on its own site.
Initial systems consultation
Choose the Salesforce MCP path before you wire it into production.
- You leave with
- A focused consultation with a senior systems consultant
- A current-state fit and architecture read
- A recommended path — implementation, ongoing ownership, or product