SF AI Toolkit

Make every Salesforce DX project AI-ready in minutes.

SF AI Toolkit is a VS Code and Cursor extension that scaffolds everything an AI coding assistant needs to work safely and effectively inside a Salesforce DX project. It generates Cursor rules, Claude Code configuration, cross-tool AI policy docs for Codex/Antigravity-style workflows, Salesforce DX MCP setup, real-time Apex and LWC diagnostics, Agentforce context detection, team config sync, and more — without overwriting existing files.

The Problem It Solves

Every Salesforce developer using Cursor or Claude Code has to manually configure MCP, write Cursor rules, set up CLAUDE.md, add skills, and figure out guardrails from scratch for every project. There is no standard. Every project ends up different.

SF AI Toolkit solves this with one command. Open a Salesforce DX project, run the setup, and your project has everything configured correctly — consistent across every developer and every project in your team.

Installation

Install from the VS Code Marketplace:

1. Open VS Code or Cursor
2. Go to the Extensions panel (Cmd+Shift+X / Ctrl+Shift+X)
3. Search for SF AI Toolkit
4. Click Install

The extension activates automatically when you open a folder containing sfdx-project.json.

Getting Started

Step 1 — Scan your project

Open the Command Palette (Cmd+Shift+P / Ctrl+Shift+P) and run:

SF AI Toolkit: Scan Salesforce Project

This scans your project and opens a readiness report showing your AI readiness score (0–100), what is missing, and what is already configured. The score also appears in the status bar at the bottom of the window.

Step 2 — Apply the setup

Run:

SF AI Toolkit: Apply Recommended Setup

Select a preset (start with core for any standard Salesforce DX project) and confirm. SF AI Toolkit creates all missing files. Existing files are never overwritten.

Step 3 — Configure MCP for your org

Run:

SF AI Toolkit: Bootstrap MCP Config

Enter your Salesforce org alias when prompted. SF AI Toolkit generates:

• .cursor/mcp.json — MCP configuration for Cursor
• .mcp.json — MCP configuration for Claude Code

Both files use the correct args[] array format. Restart Cursor or VS Code to activate.

Step 4 — Open a Salesforce file

Open any .cls, .trigger, or LWC .js/.html file. Inline diagnostics appear automatically as red and yellow underlines for detected anti-patterns. Hover over any underline for an explanation, fix suggestion, and link to the relevant Salesforce documentation.

90-Second Demo Flow

1. Run SF AI Toolkit: Scan Salesforce Project.
2. Run SF AI Toolkit: Apply Recommended Setup with core.
3. Run SF AI Toolkit: Bootstrap MCP Config.
4. Open an Apex or LWC file to show live diagnostics.
5. Run SF AI Toolkit: Check Template Drift to show team consistency workflows.

What Gets Created

Running SF AI Toolkit: Apply Recommended Setup with the core preset creates the following files if they do not already exist:

AGENTS.md                              AI tool usage guide for the project
CLAUDE.md                              Claude Code rules and workflow orchestration
AI_INSTRUCTIONS.md                     Canonical cross-tool AI policy file
tasks/todo.md                          AI task tracker
tasks/lessons.md                       Lessons learned log

.cursor/rules/
    project.mdc                        Master workflow rules (always applied)
    apex.mdc                           Apex coding standards
    lwc.mdc                            LWC coding standards
    salesforce-mcp.mdc                 MCP-first org operation rules
    deployment.mdc                     Deployment safety gates
    safety.mdc                         Security guardrails

.cursor/skills/
    11 SF AI Toolkit Salesforce skills
    29 AFV-compatible skills
    (40 total)

.claude/commands/                      6 slash commands for common workflows
.claude/agents/                        5 sub-agent definitions
docs/                                  11 reference documentation files (includes Codex + Antigravity setup)

AI Tool Compatibility

Presets

Choose the preset that matches your project type when running the setup:

Inline Diagnostics

Diagnostics appear automatically when you open supported files. No configuration required.

Apex — .cls and .trigger files

LWC JavaScript — .js files inside lwc/

LWC HTML — .html files inside lwc/

Hover over any underlined code to see the rule title, a detailed explanation, a suggested fix, and a direct link to the relevant Salesforce documentation page.

Commands

Open the Command Palette (Cmd+Shift+P) and search for SF AI Toolkit:

Team Setup

If your entire team uses Cursor on Salesforce DX projects, you can enforce a consistent AI configuration across all developers automatically.

Step 1 — Create a team config file and publish it to a URL your team can access (for example, a raw GitHub URL):

{
  "version": "1.0",
  "requiredFiles": ["AGENTS.md", "CLAUDE.md", ".cursor/rules/apex.mdc"],
  "fileSignals": {
    ".cursor/rules/apex.mdc": ["Bulkify", "with sharing"]
  }
}

Step 2 — Add the URL to your team's shared VS Code settings (.vscode/settings.json committed to your repos, or pushed via your MDM):

{
  "sf-ai-toolkit.teamConfigUrl": "https://raw.githubusercontent.com/your-org/your-repo/main/ai-kit-team.json"
}

Step 3 — SF AI Toolkit checks automatically on workspace open. If any developer's project has drifted from the team standard, a notification appears with a link to a full drift report.

Safety

SF AI Toolkit is safe to run on existing projects:

CLI

All features are available via the companion CLI without installing the extension:

npx @sf-ai-toolkit/cli scan
npx @sf-ai-toolkit/cli init --preset core --yes
npx @sf-ai-toolkit/cli bootstrap-mcp
npx @sf-ai-toolkit/cli deploy-preview
npx @sf-ai-toolkit/cli agentforce-scan
npx @sf-ai-toolkit/cli doctor

Requirements

• VS Code 1.85 or later (or Cursor)
• Node.js 18 or later (for the CLI)
• A Salesforce DX project with sfdx-project.json at the root

Author

Built by Nikhil Karkra

License

MIT