Architectural Overview
Agentforce for Developers (formerly Einstein for Developers) provides AI-powered Apex and Lightning Web Component (LWC) code generation, inline autocomplete, and unit test scaffolding inside VS Code and Code Builder. When developers authenticate their IDE to a Salesforce org, the IDE checks the org’s AgentforceForDevelopersSettings metadata (AgentforceForDevelopers.settings-meta.xml) to determine if AI coding assistance is permitted.
For regulated enterprises, controlling AgentforceForDevelopersSettings via SFDX ensures consistent AI governance across developer sandboxes, scratch orgs, and production environments.
SFDX Directory & File Suffix
| Property | Value |
|---|---|
| Metadata Type Name | AgentforceForDevelopersSettings (Settings:AgentforceForDevelopers) |
| SFDX Directory | force-app/main/default/settings/ |
| File Name & Suffix | AgentforceForDevelopers.settings-meta.xml |
| API Version Introduced | 62.0 (Verified on 68.0 / Winter ’27) |
Key Metadata Fields
| Field Name | Type | Required | Architectural Purpose |
|---|---|---|---|
enableAgentforceForDevelopers |
boolean |
No | Enables (true) or disables (false) Agentforce for Developers across the authenticated org. |
Note
Even if a developer has the Agentforce for Developers extension installed in VS Code, AI features are governed by the default org authenticated in their SFDX workspace.
Declarative XML & package.xml Examples
Sample AgentforceForDevelopers.settings-meta.xml
<?xml version="1.0" encoding="UTF-8"?>
<AgentforceForDevelopersSettings xmlns="http://soap.sforce.com/2006/04/metadata">
<enableAgentforceForDevelopers>true</enableAgentforceForDevelopers>
</AgentforceForDevelopersSettings>Sample package.xml Manifest
<?xml version="1.0" encoding="UTF-8"?>
<Package xmlns="http://soap.sforce.com/2006/04/metadata">
<types>
<members>AgentforceForDevelopers</members>
<name>Settings</name>
</types>
<version>68.0</version>
</Package>SFDX CLI Retrieval & Deployment
Retrieve and deploy AgentforceForDevelopersSettings using the sf CLI:
# Retrieve Agentforce for Developers settings from an org
sf project retrieve start --metadata "Settings:AgentforceForDevelopers"
# Deploy Agentforce for Developers settings to a sandbox or scratch org
sf project deploy start --metadata "Settings:AgentforceForDevelopers"Senior Architect Best Practices & Gotchas
- Scratch Org Default Enablement: Developers spend most of their time authenticated against scratch orgs. Include
agentforceForDevelopersSettings: { enableAgentforceForDevelopers: true }insideconfig/project-scratch-def.jsonso developers don’t lose AI assistance when switching to a newly created scratch org. - Generated Code Review Standards: While Agentforce for Developers accelerates ApexClass authoring, enforce static code analysis (PMD / Salesforce Code Analyzer) in your PR pipelines to verify bulkification and
WITH USER_MODEenforcement on all AI-assisted code.
Tip
Combine AgentforceForDevelopersSettings with well-documented Apex interfaces and clear description tags on your custom objects and fields—Agentforce uses your local SFDX schema metadata to ground its code suggestions.
Frequently asked questions
What is AgentforceForDevelopersSettings in Salesforce?
It is the org-wide Settings metadata type that enables or disables Agentforce for Developers AI code generation and inline completions for authenticated users.
Where is AgentforceForDevelopersSettings stored in an SFDX project?
It resides at force-app/main/default/settings/AgentforceForDevelopers.settings-meta.xml and is referenced in package.xml under Settings as AgentforceForDevelopers.
Why should architects manage AgentforceForDevelopersSettings in source control?
Managing this setting in Git ensures security and compliance teams have an auditable record of whether AI-assisted Apex and LWC code generation is permitted in each org.
How do I retrieve AgentforceForDevelopersSettings using Salesforce CLI?
Run sf project retrieve start --metadata Settings:AgentforceForDevelopers to download the current org configuration.
Comments
Questions, corrections and tips are welcome. Comments are GitHub Discussions, so you sign in with GitHub.