Category · Salesforce MCP servers
Salesforce DX MCP Server: Connecting AI to Your Org and Your Codebase
Salesforce DX work splits into two halves that rarely talk to each other: the metadata and code in your repository, and the live data and configuration in the org. An MCP setup is useful here precisely because it can span both — the Salesforce MCP server reaches org records, while repository and filesystem servers cover the DX project on disk. This page covers what is available in 2026 and how to assemble a workflow that sees both sides.
What this category covers
CRM and platform MCP servers connect assistants to Salesforce data and the tooling around it. The Salesforce MCP server in this directory provides access to CRM records, free with a Salesforce plan. Salesforce DX development itself is a source-control workflow — a project directory of metadata, Apex, and LWC under Git — so the servers that matter for the development half are the same ones any codebase uses: Filesystem, Git, and GitHub.
How it works
Connect the Salesforce MCP server with credentials for the org you want the assistant to reach, and add Filesystem and Git servers scoped to your DX project directory. The assistant can then answer questions that cross the boundary — reading an Apex class from the repository and checking the object schema or record data it operates on, in one conversation. That crossing is the whole value; either half alone is available in tools you already have. Our guide to the best MCP servers for coding covers the development-side picks in more depth.
Buying guide
Use read-only credentials against production and give the assistant a sandbox or scratch org if it needs to write anything. Salesforce org data is usually the most regulated data an organisation holds, and record contents pulled into a conversation are sent to your AI provider — so scope access to the objects the task actually needs rather than granting a broad integration user. On the development side, scope the filesystem server to the DX project directory and nothing above it; Salesforce projects sit alongside credential files often enough that a loose scope is a real risk. Finally, expect the assistant to be much stronger at Apex, SOQL, and LWC than at declarative configuration, where the answer usually depends on org state a repository cannot show. Teams in similar enterprise platforms should also read our ServiceNow MCP server page.
The Tools, Ranked
Connects AI agents to Salesforce CRM records. Free with a Salesforce plan and the starting point for anything touching live org data.
Scoped to your sfdx project directory, this gives the assistant access to metadata, Apex classes, and LWC components on disk.
Salesforce metadata diffs are notoriously hard to read. Repository history helps an assistant explain why a field or permission set looks the way it does.
DX changes ship through pull requests like any other code. Repository access lets the assistant read the diff and prior discussion.
Static analysis and security scanning across a codebase, useful before metadata changes reach a shared org.
Most Salesforce teams track changes in Jira, so linking a metadata change to its ticket keeps context together.
For organisations running both systems, or evaluating a migration between them.