The recent examples provided on using AI and Salesforce are disconnected.
That is, you need to do something in Salesforce, so you provide all relevant information to an AI. It gives you an answer, and you implement it. But the AI doesn’t “see” your Salesforce instance and understand the entire context. You need to provide that (if required) in your prompt.
Being disconnected can be acceptable. But it can be extremely helpful if they are connected. Imagine asking a question and the AI already understands how your Salesforce instance is configured.
This is the concept behind the new Salesforce DX MCP (Model Context Protocol), which was implemented in July 2025.
While I haven’t actually set up an AI with Salesforce, a fellow reader has. He wrote a blog post about it on Setting Up Salesforce DX MCP with Claude Code in IntelliJ
Some considerations:
- It uses the Claude AI, and you need to purchase a subscription for the Pro edition (currently 20 USD/month)
- It requires the use of Salesforce DX and CLI, which are quite technical
- You’ll need an MCP server that is specifically designed to interact with Salesforce
I can imagine these requirements will be easier to set up in the future. For now, it remains mostly accessible by technical consultants. However it’s fascinating to see that this is actually possible.
The takeaway
The lines between your preferred AI and your Salesforce instance are blurring. It’ll be interesting to see how they continue to consolidate these technologies.