Skip to main content
GPTfy - Salesforce Native AI Platform

MCP Server: Live Salesforce Data Where You Work

GPTfy turns your agent prompts into MCP tools so Claude, ChatGPT, or any MCP-enabled client can pull live Salesforce data on demand—no switching apps.

Model Context Protocol (MCP) is a universal connector: one standard way for your AI to connect to Salesforce, email, calendar, or any system. GPTfy exposes your agent prompts as MCP tools so you get live CRM data inside Claude, ChatGPT, or other MCP clients without leaving the chat.

What's inside this demo

What is MCP?

  • MCP (Model Context Protocol) is like a universal connector—one standard way for AI to connect to Salesforce, Gmail, database, calendar, and more.
  • Without MCP, you copy/paste between apps; with GPTfy MCP, your AI fetches live data on demand.

GPTfy + MCP

  • Build agents and agent prompts in Salesforce with GPTfy; those prompts become MCP tools.
  • Any MCP-enabled client (Claude, ChatGPT, etc.) can then call those tools and pull live Salesforce data into the conversation.

Demo: Account plan in Claude

  • Ask Claude for 'all information for Apple account from my Salesforce'—GPTfy MCP fetches contacts, opportunities, cases, and delivers a 360 view inside Claude.
  • Ask Claude to analyze that data and create an account plan; you get a detailed, referenced plan without leaving the chat.

Use this video when

You use Claude or ChatGPT and want live Salesforce data (accounts, opportunities, cases) inside the same chat.

You want to build account plans, forecasts, or outreach strategies using real CRM data without switching to Salesforce.

You need one standard way to connect your AI to Salesforce and other systems (MCP as universal connector).

You already have GPTfy agents in Salesforce and want to expose them to MCP-enabled clients.

Frequently asked questions

The demo shows a sales rep using Claude (the AI assistant) to pull live Salesforce data on demand and generate a complete account plan—without opening Salesforce. GPTfy turns agent prompts into MCP (Model Context Protocol) tools that Claude can call. In the demo, the rep asks Claude to retrieve all information for the Apple account, and GPTfy's MCP tools query Salesforce in the background, returning contacts, opportunities, cases, and activities. The rep then asks Claude to analyze the data and produce a structured account plan with stakeholder strategy, opportunity analysis, account health, and expansion opportunities—all referenced to live CRM data.

MCP stands for Model Context Protocol—a standard that allows AI systems to connect to external tools and data sources using a single protocol. Think of it as a universal connector: instead of building a separate integration for every AI tool, one MCP server lets any MCP-enabled client (Claude, ChatGPT, Cursor, or others) connect to your Salesforce org. For sales teams, this means you can use the AI assistant you already prefer (e.g. Claude) and have it access live Salesforce records, opportunities, and cases on demand—without switching tabs or copy-pasting data.

GPTfy MCP works with any MCP-enabled client, including Claude (Anthropic), ChatGPT (with MCP plugin support), Cursor, and other tools adopting the Model Context Protocol standard. On the Salesforce side, it works with any org where GPTfy agents are configured—Sales Cloud, Service Cloud, CPQ, and others. The MCP tools are built from the agent prompts and skills already configured in GPTfy, so there is no additional setup beyond enabling the MCP server and connecting your preferred AI client.

In GPTfy, navigate to the Agent section and enable the MCP server option. GPTfy will generate an MCP server endpoint and credentials. In your AI client (e.g. Claude Desktop), add GPTfy as an MCP server using the endpoint URL and authentication token. Once connected, Claude can see and call the GPTfy-exposed tools, which correspond to your agent prompts (e.g. get account summary, find open opportunities, retrieve cases). No code changes are needed in Salesforce; all tools derive from existing agent prompt configurations.

Yes. The MCP server authenticates every request using GPTfy's existing Salesforce OAuth flow. Data is fetched from Salesforce on-demand when the AI calls a tool, not stored in the AI provider's systems. You control which data GPTfy tools expose by configuring which agent prompts are MCP-enabled. Field-level security and record-level sharing rules are respected—the MCP tools only see data that the authenticated user is allowed to see. All tool calls are logged in GPTfy's AI Audit Log for compliance and visibility.

MCP is ideal when you need to synthesize multiple Salesforce records into a deliverable—an account plan, forecast summary, outreach strategy, or deal review—and want to do it using a conversational AI. It is also useful for users who prefer working in Claude or ChatGPT and find the Salesforce UI cumbersome for research tasks. For creating or updating records (new quotes, case updates, opportunity stage changes), using the Salesforce UI or GPTfy agents in Slack is more appropriate. MCP excels at read-heavy, analytical tasks where the AI adds value by connecting dots across many records.

Ready to see this in your Salesforce org?

Book a 45-minute session and we'll walk through this use case using your own data.

Book a Demo
Video transcript
Hello everyone, and welcome to this demo. I have some really exciting topic to talk about today. We at GPTfy have now enabled the capability to turn your agent prompts into MCP tools. So GPTfy allows the users to make agents inside Salesforce, and then you can then have agent prompts, have MCP Claude, ChatGPT or whatever MCP-enabled client you leverage. You can pull in now live Salesforce data on demand. So before jumping into the demo, I would like to take thirty seconds to tell you guys what an MCP server actually is, as that is the engine behind everything that you're about to see. So think of your AI agent like a smart employee that you have. He's sitting in a room. On its own, it can think, it can come up with answers and reasons, and give you answers. But what happens when you want to check a Salesforce record, or you wanted to check an email, or you want him to look at your particular calendar record? So it needs a way to connect to those systems, and that's exactly what an MCP server does. An MCP server stands for Model Context Protocol, and to put it in plain English, it is like a universal connector. So MCP gives your AI one standard way to connect to anything. It can be your Salesforce, it can be your Gmail, it can be your database, it can be your calendar, you name it. So one plug, endless possibility. Now let's see it in action. All right, so for this demo, I'm gonna do everything from my Claude. So this is a normal Claude window. A new chat is opened over here. And let's say I'm a sales rep and I'm new. It's my first job, it's a new team, and I don't know how exactly an account plan works. So normally how we do things these days, go to any LLM, and I'm gonna ask, "What is an account plan?" And LLM being LLM, it's gonna bring all that information to me. Okay, it's an account overview, parent relationships, opportunity maps, action plans. This is how everybody works these days. But I am new to the firm, and I have a responsibility to make an account plan for Apple. Now, that's a very big account, and so I really don't wanna mess this up. And now if I have to get access to my Salesforce data, whatever we have stored for the Apple account, I would have to stop this instance, go over there and look, download, copy/paste information. But with MCP server, I really don't have to. I can just ask my Claude, "Okay, can you bring me all the information for Apple account from my Salesforce?" As simple as that. So in the background, this instance has used our agent prompts that we have built using GPTfy, and it's gonna go inside my Salesforce, look up that Apple account, and then use prompts to find opportunities, tasks, cases, everything. It's gonna give me like a complete 360 of the Apple account right here in my Claude. And this is live Salesforce data coming into this instance. So I really don't have to jump anywhere to access this information. Okay, let's see what it has brought for me. So it has brought all the contacts and opportunities. It says eleven open cases, one closed case. I have everything. Okay. Now, I want— As I said, I don't know how to make an account plan. I'll just ask my Claude, "So can you help me analyze this data and make an account plan?" I'll just hit enter. And I'm just gonna sit back and let Claude do his thing. All right, I have my account plan. Let's see what Claude gave us using the Salesforce data I've got. Okay, account overview, executive summary, stakeholder map. Wow. Stakeholder strategy notes, opportunity analysis, opportunity deep dive. Do you see how detailed it is? Expansion opportunities, account health, and open cases. And you can see everything is referenced. Everything is coming from our Salesforce. And this solves a lot of things. This is the future. And I can just access my live Salesforce data on demand from anywhere. And as I made an account plan for this, I can use it for forecast, I can do this for generating outreach strategies for new prospects, everything on the go. If you like this demo, and if you wanna see this in action or you wanna see it for your particular cases in action, do book a demo with us and we would love to have a conversation. Thank you so much.

Explore More

Demo Videos

Browse all Salesforce and AI demo videos.

Agent Builder

Create AI agents in Salesforce with GPTfy.

AI for Sales

See how GPTfy accelerates the full sales cycle.

Last updated: March 2026