Navigate the Agentforce Vibes Landscape
Learning Objectives
After completing this unit, you’ll be able to:
- Identify the high-level product landscape of Agentforce Vibes: Skills, MCP Tools, and Rules.
- Describe where Agentforce Vibes runs and how it connects to your org.
- Explain how Agentforce Vibes keeps your org data secure through the Agentforce Trust Layer.
What Makes Up Agentforce Vibes
Agentforce Vibes is built from three core components that work together to make AI-assisted development fast, accurate, and governed. Each component plays a distinct role, and understanding them helps you get the most out of every session.
Where Agentforce Vibes Runs
Agentforce Vibes is available in two environments.
- Agentforce Vibes VS Code Extension: An extension for Visual Studio Code that brings agentic Salesforce development to your local environment. Install it from the VS Code Marketplace.
- Agentforce Vibes IDE (integrated development environment): The browser-based version of the Vibes Extension built into Salesforce sandbox orgs. No installation required. Open it directly from Setup.
Both environments connect directly to your org, giving the AI full awareness of your setup.
Because Agentforce Vibes is embedded directly in your development environment, you can generate, test, and deploy without leaving the tool. There’s no separate window to generate in and no manual copy step to move a solution into your org.
Agentforce Vibes is a paid product. Your org has a token budget that applies to all AI requests. See Agentforce Vibes Licensing, AI Models, and Consumption Monitoring for full licensing and token budget guidance.
Skills: Prebuilt Expertise for Your Org
Skills are prebuilt, reusable instruction sets that give Agentforce Vibes expert guidance for specific tasks.
When you ask Agentforce Vibes to build something, it matches your request to relevant Skills and activates only the ones that apply. For example, a Skill might encode best practices for building record pages, creating approval processes, or configuring validation rules. For developers on your team, Skills also cover Apex testing standards and Lightning Web Component (LWC) naming conventions.
Think of Skills as Salesforce Platform knowledge that the AI can access instantly. Skills encode years of Salesforce best practices, so the AI generates platform-native solutions by default rather than generic code you'd have to adapt.
Rules: Persistent Standards for Your Team
Rules are persistent instructions that apply to every session in your Agentforce Vibes environment.
You configure Rules once, and they automatically shape every request across all projects from that point forward. For example:
- All generated components must follow your team's naming conventions.
- Validation rules must include helpful error messages for end users.
- Generated test coverage must meet your org's minimum threshold.
- For developers: Apex classes must include documentation comments.
Rules are persistent and project-scoped: configure them once, and they're enforced from then on, across every conversation and every team member. AI-generated solutions stay consistent across your org without anyone having to repeat your standards in each new session.
MCP Tools: Extending What Agentforce Vibes Can Do
Model Context Protocol (MCP) is an open standard that lets AI models connect to tools and data sources in a structured, secure way.
Agentforce Vibes includes 50+ MCP tools that give the AI the ability to take actions beyond generating solutions. This means the AI can run tests, deploy changes, check quality, and validate your work without you having to manually switch between different tools or screens.
For you, MCP tools mean fewer manual steps. Instead of generating a solution in one place and then running tests in another, the AI handles the full workflow in a single conversation.
With MCP Tools, the AI becomes an active participant in your entire workflow, not just the generation step. It can run your tests, deploy to your org, and validate dependencies, so those steps happen in the same conversation instead of in separate tools you switch between by hand.
Agent Harness and AI Providers
Agentforce Vibes includes an Agent Harness that manages how the AI processes your requests. A key part of this is AI provider support, so you can choose from multiple AI models to power your development sessions.
Different models have different strengths. Some excel at complex reasoning and architecture decisions. Others are faster for straightforward generation tasks. The Agent Harness lets you switch between available AI providers so you can match the model to the task at hand.
All model requests, regardless of which provider you choose, flow through the Agentforce Trust Layer. Your data and code are never used for model training, no matter which AI provider powers your session.
For a current list of available models, see Agentforce Developer Guide: Supported Models.
Debug Logs and Trace
When you need to understand what’s happening under the hood, Agentforce Vibes provides built-in access to debug logs and trace capabilities. You can capture and review execution logs directly within your development session, without switching to a separate debugging tool.
This means when something isn’t working as expected, you can quickly pull up debug logs, trace the execution path, and identify exactly where an issue occurs, all within the same conversation where you’re building.
Guardrails and Autonomy Controls
Agentforce Vibes puts you in control of how much autonomy the AI has during your development session. You decide what gets generated and what gets applied, using the guardrails built into Agentforce Vibes.
Key guardrails include:
- Plan mode approval: When Plan mode is on, no code is generated until you review and approve the plan. This gives you a checkpoint before any changes are made.
- Human-in-the-loop by default: Agentforce Vibes proposes changes for you to review. You decide what to accept, modify, or reject before anything is applied to your org.
- Rules as safety boundaries: Rules you configure act as guardrails that shape every session. For example, you can create a Rule that generated automation must include error handling, or that code must follow specific naming conventions.
- Sandbox-first development: Agentforce Vibes works within Salesforce sandboxes, so you can experiment and iterate without risk to production data or configurations.
These guardrails mean you get the speed of AI-assisted development with the control and oversight that enterprise work requires. You set the boundaries, and Agentforce Vibes works within them.