This repository contains configuration assets, metadata, and supporting resources for Salesforce Agentforce agents. It serves as a centralized reference for defining, deploying, and maintaining trusted agentic AI within the Salesforce ecosystem.
Agentforce is Salesforce’s agentic layer — enabling autonomous, goal-oriented agents to perform tasks, engage with users, and drive productivity across Sales, Service, and Industry Clouds.
This repository provides configuration examples, reusable patterns, and trust-aligned practices to help implement and scale these agents safely and effectively.
topics/– Predefined topics and topic instructions for various use casesactions/– Standard and custom actions available to agentsguardrails/– Ethical, security, and operational guardrailstrust-layer/– Integration with the Einstein Trust Layer for monitoring and governancesamples/– Example JSON/YAML agent configurations and deployment scripts
This repository aligns with Salesforce’s Responsible Agentic AI Principles:
- Accuracy – Ground responses in Salesforce data
- Safety – Enforce prompt and scope controls
- Honesty – Disclose AI-generated content transparently
- Empowerment – Keep humans in the loop for critical decisions
- Sustainability – Use efficient, task-optimized models (e.g., xLAM, xGen-Sales)
- Clone this repository
- Review and modify configuration files under
/topicsand/actions - Deploy through Salesforce Setup → Agentforce Builder
- Monitor through the Einstein Trust Layer audit logs
- Salesforce org with Agentforce enabled
- Admin permissions to configure AI agents
- Familiarity with Topic Instructions, Actions, and Guardrails
This project follows Salesforce internal use and distribution policies.
Refer to your team’s governance and compliance documentation before deployment.
Contributions are welcome!
Please follow your org’s standard pull request process and ensure all configurations comply with internal AI Acceptable Use Policy (AI AUP) and Model Containment Policies.