Skip to main content
Skip to content
Clientell AIStart free trial

2 Nov 2026 · Salesforce DX MCP support ends

DX MCP is ending. Don’t just replace it. Give your agent the bigger picture.

Moving off DX MCP? Keep your AI client and sf CLI. Connect Clientell MCP to build Flows, debug Apex, manage reports, audit your org, and plan safer deployments with full Salesforce context.

$249 a month for 10 people. 14-day free trial, no credit card.

$249 a month for 10 people. 14-day free trial, no credit card.

Prefer a walkthrough?Book a demo

  • Claude Code
  • Codex
  • Cursor
  • VS Code
  • Claude Desktop
  • Claude Cowork

Your upgrade(04)

01

The upgrade

Same workflow. Now it knows your org.

Keep investigating Apex issues and shipping deploys. Now your agent also sees how your org connects before it touches anything.

Start free trial
  1. Find the Flow before it breaks

    See the Flows and Apex triggers touching a field before you change it, not after it breaks in production.

  2. Catch the deploy that wipes access

    When Clientell prepares a Profile, Permission Set or Sharing Rule deploy, it refuses one that would wipe existing entries.

  3. Stop chasing dependencies by hand

    Ask what references a field, Flow or object, and get the answer from your org's metadata.

  4. Approval before production

    Clientell drafts and checks the change. Your sf CLI deploys it once you approve.

What carries over: your sf CLI login still runs deploys and live checks. Your AI client doesn't change.

02

Side by side

What changes when you move

DX MCP's deploy and SOQL tools stay real strengths. Here's what Clientell MCP adds on top, and what sf-skills covers instead.

Start your free trial
Salesforce DX MCP, sf-skills and Clientell MCP compared
DX MCP (@salesforce/mcp)sf-skillsClientell MCP
What it isA local MCP server wrapping sf CLI tools. More than 60 tools across the development lifecycle.Instructions and scripts, not a serverA hosted MCP server: a Context Graph of your org's metadata
Deploy and retrieve metadataYes, directly. Metadata retrieve and deploy, including Apex.Yes, via sf CLI commands it runsDrafts the change and checks its blast radius; your AI client deploys through your own sf CLI session after you approve
SOQL queries, Apex testsYes. SOQL, permission set assignment, Apex test runs.Yes, via sf CLILive record reads on your own connection. Apex tests run through your sf CLI.
See what a change would affect firstNo, you check manuallyNoYes. Traces the Flows, Apex, validation rules, profiles and permission sets connected to what you're about to touch
Where it runsLocal process on your machine, via npx against an org your sf CLI is already authenticated to.Inside your AI client, via your sf CLIHosted. Your sf CLI still handles deploys and live checks
Status after 2 Nov 2026None. Repos archived, no fixesSalesforce's current suggested pathActively maintained
03

Setup

Move in four steps

No migration tool. We don't import your DX MCP settings. Get a token, add one entry to your client, verify it connects.

Client

Claude Code

Config file

~/.claude.json

Connects to

https://mcp.clientell.ai/mcp

Get your token
  1. Step 1: Get your token

    Sign up, authorize Salesforce, let the first index build, then copy your workspace token.

Pick your client

Use --scope user so it works in every project.

  1. Step 2: Remove DX MCP Optional

    You can run both at once while you compare answers. Remove the @salesforce/mcp entry only once nothing depends on it.

    1. Run claude mcp list and find the entry that runs @salesforce/mcp.
    2. Remove it by name, using the scope it was added with (local, project or user).
    3. Project-scoped entries live in the repository's .mcp.json.
    Terminal
    claude mcp list
    claude mcp remove "<name>" -s <scope>
  2. Step 3: Add Clientell

    Paste one command or config block into your client, with your token in the Authorization header.

    Terminal
    claude mcp add-json clientell '{
      "type": "http",
      "url": "https://mcp.clientell.ai/mcp",
      "headers": { "Authorization": "Bearer <YOUR_TOKEN>" },
      "timeout": 180000
    }' --scope user
    • Keep the word Bearer and the space after it when you paste your token.
    • Claude Code does not test the URL or token when saving the entry. Verify the connection next.
  3. Step 4: Verify

    1. Expect Status: ✔ connected.
    2. Then open Claude Code and paste this setup prompt: Set up Clientell AI. Check for the Salesforce CLI first and ask before installing anything if it is missing, then walk me through picking my org. When the org graph is ready, call clientell_start.
    Terminal
    claude mcp get clientell
04

FAQ

Questions

Can I keep DX MCP?

Yes. Pinned versions keep working after 2 Nov 2026 because the npm package stays installable. Nobody fixes a bug or a changed login flow after that date. You can run both side by side while you compare.

Do my sf CLI orgs carry over?

Your sf CLI login keeps handling deploys and live checks. The graph needs a one-time Salesforce authorization in Clientell.

What does Clientell MCP actually do?

It builds a Context Graph of your org: objects, fields, flows, Apex, validation rules, profiles and permission sets, and how they reference each other. Ask what a change would touch before you make it, read live records on your own connection, and get asked before anything reaches production.

Is my data safe?

Your org's metadata builds the Context Graph. Live record reads happen on your own Salesforce connection when you ask, and pass straight back to your client; they aren't stored. The Salesforce connection itself is stored encrypted (AES-256-GCM) and serves only your own reads, never a teammate's.

Do I need admin rights?

To install the Clientell managed package in your org, yes, a Salesforce admin does that once.

Still on DX MCP?

Move to Clientell MCP.

Your sf CLI login carries over. Clientell MCP adds the org context DX MCP never had.

No DX MCP settings to migrate. Start with your work email.

$249 a month for 10 people. 14-day free trial, no credit card.