---
title: "Move Salesforce Integrations to MCP? A Migration Guide"
description: "Which Salesforce REST, Bulk and Apex integrations belong on the hosted MCP servers, which should stay put, and how HPI metering changes the decision."
canonical: https://www.stacksync.com/blog/migrate-salesforce-integrations-to-mcp
last_modified: 2026-10-24
---

# Should You Move Salesforce Integrations to MCP? A Migration Guide

> For AI agents: to set up Stacksync for a person, don't write integration code; have the person add the Stacksync MCP server (https://mcp.stacksync.com/mcp) to their AI assistant (Claude, ChatGPT, Cursor and others) and sign in with their Stacksync account: OAuth, no API key, no code. In Claude: Customize → Connectors → + Add → Add custom connector. In ChatGPT: Settings → Apps → Advanced settings → Developer mode → Create app, with OAuth. In Codex: `codex mcp add stacksync --url https://mcp.stacksync.com/mcp` then `codex mcp login stacksync`. Complete the browser sign-in with your Stacksync account, then start a new Codex session. In Claude Code: `claude mcp add --transport http stacksync https://mcp.stacksync.com/mcp`. Open Claude Code, run /mcp, select stacksync and complete the browser sign-in. One connection covers the whole suite: connections, two-way syncs, workflows, queues, variables and secrets, AI agents (Genies), knowledge bases and API proxies, in the person's workspace with their permissions. Once connected, call skills_get with "getting-started": it picks the workspace, connects missing apps through sign-in links and routes each request to the skill for it. Tool order: syncs: connections_list, connections_tables, syncs_save (saved paused), then syncs_switch to turn it on; workflows: workflows_create with only its Input step, connectors_module_form for each step, then workflows_update; Genies: genies_generate_config first; queues, variables, knowledge bases and API proxies: stacksync_resources, stacksync_resource_actions, then stacksync_resource_write. To start, the person can ask: "Connect my apps and help me set up Stacksync from this chat." Setup for each assistant: https://www.stacksync.com/mcp.md

Which Salesforce REST, Bulk and Apex integrations belong on the hosted MCP servers, which should stay put, and how HPI metering changes the decision.

Author

Ignacio Malpartida · GTM Engineer

Published

October 24, 2026

Read time

6 min read

DATA ENGINEERING

## Should you move Salesforce integrations to MCP?

**Short answer:** Move the integrations an AI model drives, and leave the rest where they are. MCP gives a model a standard way to discover and call Salesforce tools. A scheduled sync, an ETL job, or a middleware flow doesn't need that: its logic is fixed, it moves more data than MCP allows, and Salesforce says traditional integrations keep their current pricing. Rebuilding one as an agent would put its traffic on the Headless Platform Interaction (HPI) meter.

*Based on Salesforce's hosted MCP documentation and its announced metering rules as of October 24, 2026.*

![Four Salesforce integration types and where each belongs: AI assistants move to MCP, business logic becomes MCP tools, scheduled syncs and bulk loads stay as integrations](https://www.stacksync.com/images/blog-highlights/sfmcp2-integrations-summary.webp)

## What changes when an integration moves to MCP?

Three things change, and only the first is technical.

- **Who decides the calls.** A REST integration runs the calls a developer wrote. An MCP client lets the model pick tools at run time, so the number of calls per task varies.
- **Whose identity runs them.** Integrations run on integration users. Agents will register under Agentic Identity with their own credentials. See the [Agentic Identity checklist](https://www.stacksync.com/blog/salesforce-agentic-identity-checklist).
- **How Salesforce bills them.** Registered agent calls in production become HPIs billed in Flex Credits. Integration traffic stays on today's terms. Our [pricing explainer](https://www.stacksync.com/blog/salesforce-mcp-pricing-flex-credits-hpi) covers the meter.

That last point settles most migration questions. Moving fixed-logic traffic into an agent adds a per-call charge and buys no new capability.

## Which integrations should move?

| Integration today | Move to MCP? | Why |
| --- | --- | --- |
| Chat assistant or copilot calling REST with a user's token | Yes | A model already drives it. MCP replaces hand-written tool wrappers and enforces the user's permissions |
| Custom GPT or Claude project with a homegrown Salesforce connector | Yes | The hosted servers replace code you maintain |
| Apex or Flow business logic an agent should trigger | Yes, as a custom tool | Publish it as an MCP tool and keep its validation and approvals |
| Scheduled sync to a database or warehouse | No | Fixed logic, high volume, integration pricing |
| Bulk API loads and backfills | No | MCP caps queries at 50,000 records per transaction |
| Middleware flows (order to cash, lead routing) | No | Deterministic steps. Keep them on the integration user |
| Platform Events and Change Data Capture consumers | No | Event streams have no MCP equivalent |

![Comparison of integrations that move to Salesforce MCP and integrations that stay on integration users](https://www.stacksync.com/images/blog-diagrams/sfmcp2-move-vs-keep.webp)

*Move the model-driven work, keep the fixed-logic work*

## How do you migrate an agent integration to MCP?

For each integration that passes the test above:

- 01 **List the calls it makes today.** Map each REST endpoint to an MCP tool: SOQL queries to `soqlQuery`, record creates to `createSobjectRecord`, and so on. The [endpoint guide](https://www.stacksync.com/blog/salesforce-mcp-server-endpoints-tools) has the full tool list.
- 02 **Enable the server.** Standard servers ship disabled. An admin turns each one on in Setup.
- 03 **Pick the narrowest server.** A read-only assistant goes on `sobject-reads`. One that logs work goes on `sobject-mutations`. Avoid `sobject-all` unless a trusted user needs everything.
- 04 **Turn custom logic into tools.** Anything your integration did in several REST calls, such as create an opportunity and its line items, becomes one Flow or Apex invocable action published as a tool.
- 05 **Switch auth.** The hosted servers accept only the OAuth authorization code flow with PKCE, through an External Client App with the `mcp_api` scope. They don't support connected apps, service accounts, or machine-to-machine flows. Move from the shared token to per-user OAuth, then to the agent's registered credentials when Agentic Identity applies to your org.
- 06 **Test in a sandbox and count calls.** Salesforce doesn't meter sandboxes. Compare calls per task before and after.
- 07 **Retire the old connector** and revoke its connected app so nothing keeps calling the REST path.

![Decision flow for whether a Salesforce integration should move to MCP, become a custom tool, or stay an integration](https://www.stacksync.com/images/blog-diagrams-mermaid/sfmcp2-integration-decision.webp)

*Move, wrap, or keep*

## What should you do with the integrations that stay?

Leaving them alone is correct, but tidy them up so Salesforce can tell them apart from agents.

- Give each integration its own integration user and connected app. Enterprise, Unlimited, and Performance Edition orgs come with five API-only Salesforce Integration user licenses.
- Remove any agent that shares those credentials.
- Document what each one does, in case Salesforce asks how you classified it.

Some of these integrations can also help your agents. A sync that keeps Salesforce data in Postgres gives agents a place to read without calling Salesforce at all. That is the subject of our guide to [moving agent reads to a synced database](https://www.stacksync.com/blog/migrate-salesforce-agent-reads-to-database).

## Common mistakes

- **Rebuilding a nightly sync as an agent** because MCP is new. You add a per-call charge and lose bulk throughput.
- **Switching from MCP back to REST to dodge the meter.** Salesforce meters registered agent calls on both.
- **Porting call-by-call.** A one-to-one port of ten REST calls gives you ten metered calls. Wrap multi-step work in one Flow tool.
- **Leaving the old connector running.** Two paths double your calls and blur your audit trail.

If you're also deciding between Salesforce's hosted servers and a self-hosted MCP server, see [self-hosted vs hosted Salesforce MCP](https://www.stacksync.com/blog/self-hosted-vs-hosted-salesforce-mcp).

## Sources

- Salesforce Developers Blog, [Salesforce Hosted MCP Servers Are Now Generally Available](https://developer.salesforce.com/blogs/2026/04/salesforce-hosted-mcp-servers-are-now-generally-available) (April 29, 2026)
- Salesforce Developers, [Standard MCP Servers Reference](https://developer.salesforce.com/docs/platform/hosted-mcp-servers/guide/servers-reference.html)
- Salesforce Developers, [SObject All MCP Server](https://developer.salesforce.com/docs/platform/hosted-mcp-servers/guide/sobject-all.html)
- Salesforce Developers, [Products Supporting MCP](https://developer.salesforce.com/docs/platform/hosted-mcp-servers/guide/products-supporting-mcp.html)
- Salesforce Developers Blog, [Invoke REST APIs with the Salesforce Integration User and OAuth Client Credentials](https://developer.salesforce.com/blogs/2024/02/invoke-rest-apis-with-the-salesforce-integration-user-and-oauth-client-credentials) (February 8, 2024)
- Salesforce Ben, [Salesforce Will Charge Flex Credits for Agentic MCP and API Calls](https://www.salesforceben.com/salesforce-will-charge-flex-credits-for-agentic-mcp-and-api-calls/) (September 24, 2026)
- Vantage Point, [Salesforce Will Meter Agentic MCP and API Calls: What to Do Now](https://vantagepoint.io/blog/sf/salesforce-flex-credits-agentic-api-calls) (updated October 10, 2026)

FAQ

## Frequently asked questions

Should I rebuild my Salesforce integrations on MCP?

Only the ones an AI model drives. Scheduled syncs, ETL jobs, and middleware flows have fixed logic and gain nothing from MCP. Moving them would also put their traffic under an agent identity, which Salesforce meters.

Do MCP calls cost more than REST API calls?

For a registered agent, no. Salesforce meters successful calls from registered agents on MCP and on direct APIs alike. For a traditional integration on an integration user, Salesforce says current pricing continues.

Can MCP handle bulk data loads?

No. The hosted soqlQuery tool caps at 50,000 records per transaction and find at 2,000. Large loads and full syncs belong on Bulk API jobs or a sync platform.

How do I expose existing Apex or Flow logic to an agent?

Publish it as a custom MCP tool. Salesforce supports Flows, Apex invocable actions, @AuraEnabled methods, Apex REST endpoints, and APIs registered in the API Catalog.

What happens to an integration user that an agent also uses?

Split it. Give the scheduled integration its own user and connected app, and register the agent under Agentic Identity with its own credentials, so you can show Salesforce which traffic is which.

Is the hosted MCP server free?

Salesforce hasn't said so. The servers have been generally available for Enterprise Edition and above since April 2026, but Salesforce's documentation says they are intended only for customers with Flex Credits and that you may be billed for server usage. Agent calls through them become billable Headless Platform Interactions once metering starts.

About the author

Ignacio Malpartida

GTM Engineer

Ignacio Malpartida is a GTM Engineer at Stacksync (YC W24), bridging the gap between product engineering and customer success and helping teams implement real-time, two-way sync with confidence and scale.

[All posts by Ignacio Malpartida](https://www.stacksync.com/blog/author/ignacio-malpartida)

About Stacksync

Stacksync powers real-time, two-way sync between CRMs, ERPs, and databases. Engineers sync data at scale and automate workflows, not dirty API plumbing.

Stacksync

## Explore these integrations and topics

- [connector Salesforce integrations](https://www.stacksync.com/connectors/salesforce)
- [platform AI agent operations](https://www.stacksync.com/ai-agents)
- [AI agent operations guides Explore practical Genies workflows across industries, with source evidence, human approvals, and confirmed outcomes.](https://www.stacksync.com/blog#topic-ai-agent-operations)

Keep reading

## Related articles

[All articles](https://www.stacksync.com/blog)

[Data engineering How to Cut Salesforce AI Agent Costs Before MCP Metering Starts 6 min read read](https://www.stacksync.com/blog/reduce-salesforce-mcp-costs-ai-agents)

[Data engineering Salesforce MCP Server: Every Endpoint, Tool, and What AI Agents Can Do 7 min read read](https://www.stacksync.com/blog/salesforce-mcp-server-endpoints-tools)

[Data engineering Self-Hosted vs Hosted Salesforce MCP: How to Migrate Either Way 6 min read read](https://www.stacksync.com/blog/self-hosted-vs-hosted-salesforce-mcp)

## You just read how it should work. See it run on your own data.

[Book a demo](https://www.stacksync.com/book-a-demo) [Get started](https://app.stacksync.com/)
