Salesforce developers can adapt Claude Code through Salesforce’s official Development plugin, which bundles Skills, MCP servers, agents, hooks, commands and related tools. “Claude Code Mods” is not established in Salesforce’s documentation as the name of a separate feature; here, “mods” means these documented customization mechanisms.
What the Salesforce Development plugin adds
The official Salesforce Development plugin packages Salesforce-focused instructions and tools, detects Salesforce DX project and connected-org context, and can work with Salesforce CLI and Salesforce-hosted MCP services. It is a bundle, not a single switch that changes the agent’s behavior in one way. Its capabilities come from several components with different purposes. Salesforce’s plugin overview describes the bundle and its context-aware behavior.
As an Amazon Associate I earn from qualifying purchases.
| Mechanism | What it contributes | How it is triggered | Scope and review point |
|---|---|---|---|
| Skills | Reusable Salesforce-specific instructions and assets, including patterns for Apex, Flow and Lightning Web Components. | Claude Code reads skill descriptions first, then loads full instructions when a task matches. | Task guidance; check whether the relevant Skill applies to the request. |
| MCP servers | Connections that expose context, tools or services. The plugin configures salesforce-api-context, salesforce-metadata-experts and salesforce-lsp. |
Used when the agent needs the connected service or context. | Plugin-provided connections; other MCP servers are not part of the core install. Headless 360 MCP Server (Beta) requires org activation and OAuth configuration. |
| Hooks | Event-triggered shell actions, such as checking Agent Script syntax after .agent edits or validating metadata deployments. |
An event, rather than a conversational request, triggers the configured action. | Automation around agent activity; a hook is not a guarantee that generated work is correct. |
| Agents and commands | Specialized workflows and callable capabilities included in the plugin. | Used as appropriate to the task and plugin workflow. | Recommendations for specialized plugins require developer review and do not install them automatically. |
The mechanisms are complementary: Skills provide task instructions, MCP connects tools or data, and hooks automate checks around events. The plugin’s agents and commands add workflows, while project files can supply local conventions.
What you need before installing
Salesforce’s current installation guide lists these prerequisites. Versions and entitlements can change, so consult the linked guide when setting up.
#1 Best Overall
- Claude Code 2.1.222 or later.
- Node.js LTS.
- Salesforce CLI.
- Python 3.8 or later.
- Salesforce Foundations or an Agentforce license for direct plugin access.
Install and authenticate Claude Code according to Anthropic’s setup instructions, which cover supported operating systems, authentication and updates. The Salesforce plugin prerequisites are separate from Claude Code’s own installation requirements.
Install the plugin and check project readiness
- Confirm that Claude Code and the Salesforce prerequisites are installed, and that your Salesforce account has the required entitlement.
- Install the marketplace plugin using its official name:
salesforce-development@claude-plugins-official. Use Salesforce’s current installation guide for the exact installation flow. - Open Claude Code from the root of the Salesforce DX project you intend to work on, then use
/pluginto check that the plugin is installed. - Start with the plugin’s readiness prompt: “Am I ready for development?” To check deployment access, ask: “Validate that I can deploy to this org.” These are examples from Salesforce’s usage guide.
- Review the project and org context Claude Code identifies before asking it to make changes. Confirm the target org, especially before any deployment or other org action.
A successful readiness response is useful context, not permission to skip review. Ask the agent to explain the planned change and inspect the proposed actions before approving them.
Give a task with a clear outcome
Specific requests help distinguish the desired result from the steps the agent may propose. Salesforce’s usage guide demonstrates this prompt: “Generate a custom object Project__c with fields for Name, Status, Due Date, and Owner.” The guide also shows tasks such as generating metadata, writing an Apex test and deploying changes to a sandbox.
For a real project, add constraints that matter: the intended org or sandbox, applicable naming or code conventions, what tests to run, and whether the request is to prepare a change or deploy it. Inspect generated files and any proposed CLI, shell or org actions; approve only actions whose target and effect you understand.
Use Skills and project context for local conventions
Skills are the reusable instruction layer. The agent initially reads their descriptions and loads complete instructions when a task matches, so a Salesforce-specific Skill can guide relevant work without requiring the full instructions to be loaded for every request. The plugin includes guidance for Salesforce technologies such as Apex, Flow and Lightning Web Components. Salesforce’s Agent Skills documentation explains the mechanism.
Project context is distinct from a plugin Skill. Salesforce says generated Multi-Framework projects include an AGENT.md file at the project root with project conventions. That does not mean every Salesforce DX project already has one. If your project uses such a file, review its contents and keep conventions current; otherwise, provide relevant requirements in your prompt or through the project’s established configuration.
When to use MCP servers, hooks or plugin discovery
MCP for connected context and services
The Salesforce plugin automatically configures its three core MCP servers: salesforce-api-context, salesforce-metadata-experts and salesforce-lsp. Do not assume an arbitrary MCP server is included just because the plugin is installed. Salesforce identifies Headless 360 MCP Server as a beta option that requires activation in the org and OAuth configuration. Before connecting additional services, review what context and actions they expose and how access is authorized.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Hooks for repeatable event checks
A hook can run a shell action when a configured event occurs. Salesforce’s examples include detecting Salesforce DX project and org context, checking Agent Script syntax after changes to .agent content, and validating metadata deployments. Hooks can make a check repeatable, but their value depends on what they check and whether they run successfully. Review shell commands and their permissions before enabling them; do not treat a hook as proof that all code or deployments are safe.
Agents and commands for specialized workflows
The plugin bundles agents and commands, and its overview says it can recommend specialized plugins based on project signals, task prompts or manual discovery. A recommendation is not an automatic installation. Inspect what a suggested plugin contributes and approve installation intentionally.
Review permissions and deployment actions
Claude Code’s command-line options include controls for MCP configuration, model selection and allowed or disallowed tools. These are configuration choices in your environment, not guarantees made by the Salesforce plugin. Consult Anthropic’s CLI reference and use permissions appropriate to the task.
- Keep the agent in the intended Salesforce DX project so it has the right local context.
- Check which org is connected before requesting or approving org-facing work.
- Read proposed shell commands, file edits and deployment steps before approval.
- Use a sandbox when that is the appropriate target, and review test and deployment results rather than assuming a successful command validates the whole change.
- Inspect any hook or additional MCP configuration for its actions, access and failure behavior.
Salesforce describes generated changes as subject to review before they are applied. That review remains important even when a Skill supplies guidance or a hook performs a validation step.
Free tools Windows power users keep installed
One-click scans. No signup required.
What “Claude Code Mods” means here
The Salesforce documentation supports a practical customization model based on plugins, Skills, MCP servers, hooks, agents, commands and project context. It does not establish “Claude Code Mods” as an official feature name or “Moddable Agent” as a formal Salesforce product. Use those phrases as informal shorthand, not as names for an undocumented extension system.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
