Salesforce MCP: The Official Hosted Server, Explained
Yes — Salesforce has an official MCP server, hosted by Salesforce itself. The Hosted MCP Server, generally available since April 2026, lets any MCP-compatible AI tool read and write your records, run flows and Apex actions, and query your org — all through the open Model Context Protocol standard. A separate Salesforce DX MCP server covers developer workflows. So if you’re searching “Salesforce MCP,” the connection you want is real, and Salesforce built it.
The thing worth knowing before you set it up: an MCP server hands your CRM to an AI inside a conversation you start. It’s a doorway, not a worker. Nothing watches your pipeline for you, nothing fires when an opportunity changes stage, and nothing runs while the chat is closed. Here’s exactly what the Salesforce MCP does, how to turn it on, where it stops — and what to use when you want Salesforce work that runs on its own.
Salesforce is one of many — the complete list of MCP servers covers every app that ships one, grouped by category.
What the Salesforce MCP server does
Model Context Protocol (MCP) is the open standard that lets an AI client — Claude, ChatGPT, Cursor, and others — talk to an outside app through a shared interface. Salesforce maintains two official servers:
- Hosted MCP Server (GA since April 2026, Enterprise Edition and above) — Salesforce-managed endpoint that exposes your org’s data, flows, Apex Invocable Actions, @AuraEnabled methods, and Named Queries to any MCP-compatible AI client. Authentication is per-user OAuth 2.0 with PKCE, so an agent only ever acts within the requesting user’s own permissions — CRUD, field-level security, and sharing rules all still apply.
- Salesforce DX MCP Server (developer tool, open source at
github.com/salesforcecli/mcp) — built for developers running common dev tasks like deploying code, creating scratch orgs, and running tests through agentic tools like Claude Code or Cursor.
For most people the first one is what “Salesforce MCP” means. With it connected, an AI client can:
- Look up records — “pull the Acme opportunity and its related contacts” answered from the live org.
- Query and filter — “which opportunities over $50k have no activity in 2 weeks?” run against real data, not a guess.
- Create and update — add a lead, move an opportunity stage, log a case without opening Salesforce.
- Run flows and Apex actions — trigger the same automations and custom logic your admins already built, from inside a chat.
It’s genuinely useful for ad-hoc work: ask a question, get an answer grounded in your org, make a change on the spot.
How to set up the Salesforce MCP server
The Hosted MCP Server is Salesforce’s no-code path, since Salesforce hosts and scales it the same way it does your REST APIs:
- In Setup, enable the Hosted MCP Server for your org (Enterprise Edition or above).
- Create an External Client App with the scopes you want the AI to have — this is what per-user OAuth authorizes against.
- In your AI client’s connector settings, add the Salesforce MCP endpoint and complete the OAuth prompt.
- Confirm the tools appear in the client, then start a chat and ask it to read or update a record.
Salesforce’s own setup docs put the whole flow at roughly 30 minutes. The DX MCP server runs differently — you add it to your agentic editor’s MCP config from the open-source salesforcecli/mcp repo — but that path is for developers, not day-to-day CRM users.
Where the Salesforce MCP stops
None of this is a knock on MCP — it’s just the shape of the protocol. Four limits show up the moment you want more than a conversation:
- It only works inside a chat you start. Close the window and nothing happens. The AI doesn’t watch your org; it waits for you to ask.
- No triggers. An opportunity moving to Closed Won, a new lead coming in, a case going stale — none of these can start anything through MCP. There’s no “when this happens in Salesforce, do that.”
- It’s one app at a time. The Salesforce MCP knows Salesforce. Getting a won deal into Stripe, a Slack channel, and a finance sheet means wiring up (and authing) a separate MCP server for each, then hoping your client can juggle them in one turn.
- You own the plumbing and the scopes. The External Client App, OAuth tokens, and the blast radius of read/write access to your org are all on you to configure and maintain.
So the Salesforce MCP is a great way to ask your org things and make one-off edits. It is not a way to make Salesforce run — to have work happen on a schedule or in reaction to an event, across the other tools an opportunity or case touches.
Running Salesforce work that doesn’t need a chat open
That “run on its own, across apps” gap is exactly where Carly fits. Carly connects to Salesforce natively — no MCP server to enable, no External Client App to configure — and to the ~260 other apps it supports, plus anything with a public API through your own key. The difference from MCP is the important part: Carly’s workflows are triggered and scheduled, so Salesforce work happens whether or not anyone has a chat window open.
A few things that MCP can’t do but a Carly workflow can:
- When an opportunity hits Closed Won → create the Stripe customer, post to the
#winsSlack channel, and add a row to the finance sheet — automatically, the moment it happens. - Every morning → summarize opportunities with no activity in 7 days and send the list to the rep who owns them.
- When a lead comes in from a form → enrich the record, route it to the right rep, and draft the first follow-up for approval.
The non-AI steps — the moving, matching, and routing between apps — are free and unlimited, the Zapier-style backbone of the workflow. The AI steps (drafting, summarizing, deciding) start at $35/month. You describe the outcome in plain language and Carly wires up the Salesforce connection and everything downstream.
If you just want to interrogate your org from a chat, Salesforce’s official Hosted MCP Server is the right tool. If you want Salesforce to actually do things — on a trigger, on a schedule, across every app an opportunity flows through — that’s the job MCP wasn’t built for, and it’s the one Carly was.
FAQ
Does Salesforce have an official MCP server?
Yes. The Hosted MCP Server, generally available since April 2026 for Enterprise Edition orgs and above, gives MCP-compatible AI tools read/write access to your data, flows, and Apex actions. A separate open-source Salesforce DX MCP server (github.com/salesforcecli/mcp) supports developer workflows.
Is the Salesforce MCP server free? The Hosted MCP Server is included for orgs on Enterprise Edition and above — you’re authorizing an AI client against your existing org and permissions, not paying extra for the connector itself. You still need the underlying Salesforce edition it requires.
Can the Salesforce MCP trigger automations? No. MCP is request/response inside an AI chat — it has no triggers and nothing runs when the conversation is closed. For event- or schedule-driven Salesforce work across apps, you need a workflow tool like Carly rather than an MCP server.
What AI tools can connect to Salesforce over MCP? Any MCP-compatible client — Claude, ChatGPT, Cursor, and custom agents built on the open protocol — can connect to the Salesforce Hosted MCP Server.
Can I connect Salesforce to AI without coding or hosting a server? Yes. You don’t have to touch MCP at all. Carly connects to Salesforce for you and lets you build the automation in plain language — describe what you want to happen and it wires up the CRM and the other apps involved, with no server to host and no code to write.
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
