AI Agent Studio

Build intelligent AI agents powered by Large Language Models that seamlessly integrate with your Salesforce environment.

Get Started View on GitHub
GitHub Stars GitHub Forks Salesforce API License
📦 Repository Notice

This repository contains the core AI Agent Framework only. The aiAgentStudioAddons folder contains proprietary extensions including additional orchestrators, LLM providers, actions, and UI components that are not part of the open-source repository.

Why AI Agent Studio?

  • No Code Required — Configure AI agents with point-and-click tools
  • Enterprise Security — Automatic CRUD, FLS, and sharing rule enforcement
  • Multiple AI Providers — OpenAI included, extensible for Claude, Gemini, and more
  • Full Observability — Every interaction logged for transparency and debugging
  • Open Source — Free forever under MPL-2.0 license

Who is this for?

⚙️

Admins

Configure AI agents without code using point-and-click tools

📊

Managers

Understand what AI agents can do for your team

💻

Developers

Extend the framework with custom actions and integrations

Three Types of AI Agents

💬 Conversational Agents

Interactive chat assistants with multi-turn conversations and memory. Perfect for customer support, sales assistance, and help desks. Included in core framework.

⚡ Function Agents

Single-task specialists for one-shot operations like summarization or classification. Fast, stateless execution.

🔄 Workflow Agents

Multi-step automation that orchestrates complex processes with multiple agents. Sequential execution with checkpoints.

Key Features

🧠 Smart Memory

Buffer window and summary-based conversation history management

🔒 Built-in Security

Automatic CRUD, FLS, and sharing rule enforcement

🤖 Multiple AI Providers

OpenAI included, more in addons, extensible adapter pattern

🛠️ Standard Actions

Create/update records, Chatter, flows, and more

⚙️ Async Processing

Platform Events or Queueables for your needs

📊 Deep Observability

Full logging of inputs, outputs, and tool executions

Quick Start

TL;DR

Deploy the framework → Configure Named Credentials with your AI API key → Create an LLM Configuration → Create an Agent → Add capabilities → Drop the chat component on a page → Start chatting!

⏱️ Time to First Agent

You can have your first AI agent running in under 30 minutes!

Prerequisites

  • Salesforce org (Sandbox recommended)
  • System Administrator access
  • API key from an AI provider (OpenAI, Claude, or Gemini)
1

Deploy the Framework

sf project deploy start -d force-app/main/default -o your-org-alias
2

Set Up AI Provider Authentication

Navigate to Setup → Named Credentials → New

  • Label: OpenAI API
  • URL: https://api.openai.com
  • Add your API key as authentication
3

Create LLM Configuration

App Launcher → LLM Configurations → New

  • Developer Name: OpenAI_GPT4o
  • Named Credential: OpenAI_API
  • Provider Adapter Class: OpenAIProviderAdapter
  • Default Model: gpt-4o-mini
  • Temperature: 0.7
  • Is Active: ✓
4

Create Your First Agent

App Launcher → AI Agent Definitions → New

  • Name: Sales Assistant
  • Agent Type: Conversational
  • LLM Configuration: OpenAI_GPT4o
  • Memory Strategy: BufferWindow
  • Identity Prompt: “You are a helpful Salesforce assistant.”
5

Add the Chat Component

  • Edit any Lightning page
  • Drag aiAssistantChat component to the page
  • Configure with your agent’s developer name
  • Save and activate
6

Start Chatting!

  • “Find contacts named John Smith”
  • “Show me all accounts in California”
  • “Create a task to follow up with Acme Corp”
📖 Need more details?

Check out the complete Getting Started guide for step-by-step instructions with screenshots.

Standard Actions

Action Description
ActionGetRecordDetails Retrieve and search Salesforce records
ActionCreateRecord Create any Salesforce record
ActionUpdateRecord Update existing records
ActionPostChatter Post to Chatter feeds
ActionFlowHandler Execute Salesforce Flows

Additional actions are available in the addon package.

View all actions →

Security

🔒 Security First

Agents always run in the user's context. There is no privilege escalation — agents can only access what the user can access.

The framework enforces security at every level:

  1. User Context — Agents run as the requesting user
  2. CRUD Checks — Validates object-level permissions
  3. FLS Enforcement — Validates field-level permissions
  4. Sharing Rules — Respects record-level access
  5. Audit Trail — All actions logged in AgentDecisionStep__c

Learn more about security →

License

Licensed under the Mozilla Public License 2.0.

  • ✅ Commercial use allowed
  • ✅ Modification allowed
  • ✅ Distribution allowed
  • ⚠️ Must disclose source if distributing modifications

📚 Continue Learning