SidebarSite navigation.

Cert++

Practice
Certle
Review
Tracks
Study Guide

More

About
Articles
Community
Support
Changes
Legal

Agent API versus other Agentforce APIs and SDKs

Key Takeaway

Use the Agent API when code needs to talk to an Agentforce agent directly through a REST API. It doesn't build or test agents. Service Cloud web chat, mobile apps, and flows or Apex classes have their own tools.

The Agent API is a REST API that lets code talk to an Agentforce agent directly. Through it, code starts a session, sends and receives messages, and ends the session. The Agent API is one of several Agentforce developer tools. The other tools build agents, test agents, or let you talk to agents in Service Cloud chat channels, in mobile apps, and from flows or Apex classes. For the exam, the key skill is knowing when the Agent API fits and when another tool does.

Three jobs: build, test and use

Agentforce developer tools do three jobs: they build agents, test agents or use agents. Using an agent means talking to it. The Agent API is a use tool, because it lets you chat with agents through a REST API. It does not build or test them.

Agentforce Specialist Study Guide

Platform Administrator
Platform App Builder
Platform Foundations
Platform Developer
Platform Administrator II
Agentforce Sales Consultant
Agentforce Service Consultant
Platform Data Architect
Platform Development Lifecycle and Deployment Architect
Platform Identity and Access Management Architect
Platform Integration Architect
Platform Sharing and Visibility Architect
Heroku Architect
B2C Solution Architect
Experience Cloud Consultant
Agentforce Field Service and Operations Consultant
Agentforce Nonprofit Consultant
Data 360 Consultant
Omnistudio Consultant
CRM Analytics and Einstein Discovery Consultant
Platform User Experience Designer
Platform Strategy Designer
B2C Commerce Developer
JavaScript Developer
Omnistudio Developer
Platform Developer II
Marketing Cloud Engagement Administrator
Marketing Cloud Engagement Specialist
Marketing Cloud Engagement Consultant
Marketing Cloud Account Engagement Specialist
Marketing Cloud Next Consultant
Agentforce Sales Foundations
Business Analyst
Marketing Cloud Engagement Developer
Marketing Cloud Engagement Foundations
Agentforce Specialist
Agentforce Life Sciences Consultant
B2B Commerce Administrator AP
B2B Commerce Developer AP
Agentforce Communications AP
Agentforce Consumer Goods AP
Agentforce Financial Services AP
Agentforce Health AP
Agentforce Manufacturing AP
Public Sector Solutions AP
Revenue Cloud Consultant
MuleSoft Integration Foundations
MuleSoft Developer
MuleSoft Developer II
MuleSoft Platform Integration Architect
MuleSoft Platform Architect
Tableau Desktop Foundations
Tableau Data Analyst
Tableau Consultant
Tableau Server Administrator
Tableau Next Consultant
Tableau Architect
Slack Administrator
Slack Consultant
Slack Developer

Checklist progressProgress

0/204Learned

Checklist progressProgress

0/204Learned

When to use Prompt Builder
Prompt Builder versus predictive AI
Prompt Builder versus Flow automation
Prompt Builder versus Agent Builder
Prompt templates as Agentforce actions
Prompt Builder enablement and user assignment
Prompt Template Manager versus Prompt Template User
Type-specific permissions for prompt templates
Running-user data access for prompt templates
Data 360 permissions for retriever-grounded prompt templates
Field Generation prompt templates
Flex prompt templates
Sales Email prompt templates
Record Summary prompt templates
CRM merge fields for prompt grounding
Flow versus merge fields for prompt grounding
Apex merge fields versus Flow for prompt grounding
Knowledge RAG grounding in prompt templates
Data 360 retriever grounding in Prompt Builder
Combining merge fields with retrieval grounding
Web grounding for real-time external data
File input grounding in Prompt Builder
Record field merge fields versus Record Snapshots
Flow and Apex input compatibility with prompt templates
Retriever prefilters for scoping search results
Prompt template creation in Prompt Builder
Prompt template preview and revision
Prompt template activation and draft status
Prompt template execution surfaces
Customizing Salesforce-provided standard prompt templates
Prompt template anatomy and design
Merge fields versus hardcoded values in prompt text
Grounded-response instructions in prompt templates
Context engineering versus prompt engineering
Context clash, confusion, and poisoning
Consistent writing style and plain-language wording in prompts
Separating instructions from context and input data in prompt templates
Restricting LLM output to the expected content type
Choosing zero-shot, few-shot, or chain-of-thought prompting
Diagnosing and fixing weak prompt template output
Einstein Trust Layer architecture
Secure data retrieval for prompt grounding
Pattern-based and field-based LLM data masking
Prompt defense system policies
Zero-data retention with third-party LLMs
Toxicity detection on generated responses
Generative AI audit trail and feedback data
LLM data masking for agents versus embedded generative AI features
Supported LLMs for Prompt Builder and Agentforce
Org-level model enablement and blocking
Prompt template model selection
Agentforce model option and subagent overrides
Hiding an LLM configuration from Prompt Builder model selection

Purpose of Agentforce Data Library
Data Library source types
Knowledge grounding with a Data Library
Data Library setup versus runtime
Data Library setup considerations
Assigning a Data Library to an agent
Default versus additional Data Libraries
Custom retrievers in a Data Library
Data Library versus advanced RAG setup
File-based data libraries and Intelligent Context processing
Retrieval augmented generation in Data 360
Chunking sources for AI retrieval
Search indexes for chunked Data 360 content
Vector versus hybrid search indexes
Retrievers versus search indexes
Retriever creation and activation for grounding
Ensemble retrievers in Data 360
Knowledge retrieval troubleshooting for agents
Easy Setup versus Advanced Setup for search indexes
Choosing a chunking strategy for the content type
Max token and overlap token settings for chunk size
Prepend fields to add context to chunks
Enriched indexing of content chunks
Default retrievers versus custom retrievers
Individual retriever filters and returned fields
Retriever citations for verifying grounded responses

Explain how an agent works and its basic building blocks of agent script

0/8

Agentforce agent building blocks
Atlas reasoning engine request handling
Subagents versus actions
Action chaining and subagent transitions
Agent Script as the Agentforce Builder language
Agent Script block types
Next-generation authoring of Agent Script
Final response validation for Service agents

Explain the components and benefits of hybrid reasoning, including how Agent Script functions in Canvas and Script View

0/6

Hybrid reasoning components in Agent Script
Benefits of hybrid reasoning
Agent Script in Canvas View
Agent Script in Script View
Canvas versus Script View
Before and after reasoning in Agent Script

Given a use case, manage deterministic behavior for the agent using mechanisms like filters, variables, and template expressions.

0/8

Topic filters for request gating
Action filters for business-rule gating
Input variables versus hardcoded action values
Output variables for action-to-action data
Context and conversation variables
Template expressions in agent instructions
Conditional logic for context-specific prompts
Modality-aware responses for text and voice conversations

Given a scenario, select and configure standard topics, custom topics, standard agent actions, and custom agent actions.

0/14

Standard versus custom topics
Topic classification descriptions and routing
Custom topic scope and instructions
Standard versus custom agent actions
Custom action types from Flow, Apex, and prompt templates
Creating custom agent actions
Agent action descriptions and instructions
Topic-level versus action-level instructions
Assigning actions to a topic
User access for standard agent actions
Asset library versus in-agent creation of subagents and actions
Customizing a standard subagent within one agent
Customizing standard agent action behavior
When to split, combine or add subagents

Explain the process for connecting agents to various channels such as digital experience, email, voice, and Slack.

0/9

Supported channels for Employee and Service agents
Service Agent connections to digital experiences
Service Agent connections to messaging channels
Service Agent connections to email
Service Agent connections to Agentforce Voice
Omni-Channel routing to a Service Agent
Employee Agent connections to Slack
Employee Agent connections to Lightning Experience and Mobile
Agent connections, connection instructions and adaptive response formats

Explain the security context in which the agent is actually running, and how it impacts agent action execution

0/7

Execution identity by agent type
Running-user permissions and sharing for actions
Employee agent access for end users
Credential-based verification and Service run-as
Identity verification in sensitive agent actions
Context variables versus data access control
Flow run context within agent action execution

Given a scenario, identify when to use an Employee or Service agent.

0/2

Employee versus Service agent selection
Agentforce Default versus Employee agent

Given a scenario, identify when it's appropriate to use Agent API

0/6

When to use Agent API versus standard channels

Learn this concept

Unseen

Agent API supported agent types

Learn this concept

Unseen

Agent API versus Models API

Learn this concept

Unseen

Agent API versus custom agent actions

Learn this concept

Unseen

Agent API versus other Agentforce APIs and SDKs

Learn this concept

Unseen

Agent API actions in Flow and Apex

Learn this concept

Unseen

Agent testing strategy from preparation through retest
Agentforce Builder versus Testing Center testing
Builder Simulate versus Live Test modes
Permissions for Testing Center tests
Testing Center test case criteria
Test data and conversation context for Testing Center
Batch testing and iteration in Testing Center
AI-generated tests in Testing Center
Automating agent tests with Testing API and Agentforce DX
Builder validation checks and trace previews for troubleshooting
Testing Center subagent and action evaluation
Testing Center response evaluation
Testing Center results interpretation
Quality metric scorers in Testing Center
Custom Testing Center scorers
Testing Center limits and considerations
Prompt template batch evaluation scorers
Change sets versus Agentforce DX for agent deployment
Dependent components in an agent change set
Target-org prerequisites for a deployed agent
Agent activation after production deployment
Post-deployment channel and connection configuration
Agent versioning for uninterrupted production updates
Data 360 and Data Library considerations for agent promotion
Agent user reassignment and permissions after a cross-org deployment
Prompt template change sets and metadata deployment
Dependent metadata for prompt template deployment
Prompt Builder enablement in the production org
Prompt template activation in the destination org
Prompt template version behavior on deploy
Field generation Lightning page assignment after deploy
Retriever and Knowledge grounding after template deployment

Monitoring in the Agent Development Lifecycle
Agent manager permission sets
Activate and deactivate agents
Agent versioning lifecycle
Agentforce guardrails and trust patterns
AI agent risk identification
Agent guardrail monitoring
Einstein Audit and Feedback setup
Audit trail and toxicity review
Enhanced Event Logs for agent monitoring
Investigating agent sessions and intents
Omni-Channel monitoring for Service agents
Permissions for managing agents versus viewing observability data
Agent Health Monitoring alerts for silent agent failures
Agent Analytics versus Agent Optimization
Agent Analytics metrics dashboards and session insights
Session Tracing Data Model for observability
Intents in Agent Optimization
Utterance analysis for agent conversations
Quality scores and subagent performance trends
Scorers and custom scorers in Agentforce Studio
Context engineering for agent optimization
Legacy Agentforce Analytics versus Agent Analytics

Single-agent versus Multi Agent architecture
Subagents versus a Multi Agent architecture
Orchestrator routing to connected subagents
Handoff mode versus supervisor mode orchestration
Agent Script for deterministic multi-agent routing
Single-org versus multi-org multi-agent architecture (SOMA and MOMA)
When to use Model Context Protocol (MCP)
When to use Agent-to-Agent (A2A) protocol
MCP server, client and host roles
Exposing Salesforce capabilities to external AI clients with Hosted MCP Servers

Prepare for the Exam

Play Today's Certle

Study Community

Ask questions and get the latest info from other Agentforce Specialist studiers.

Agent API versus other Agentforce APIs and SDKs

Key Takeaway

Use the Agent API when code needs to talk to an Agentforce agent directly through a REST API. It doesn't build or test agents. Service Cloud web chat, mobile apps, and flows or Apex classes have their own tools.

The Agent API is a REST API that lets code talk to an Agentforce agent directly. Through it, code starts a session, sends and receives messages, and ends the session. The Agent API is one of several Agentforce developer tools. The other tools build agents, test agents, or let you talk to agents in Service Cloud chat channels, in mobile apps, and from flows or Apex classes. For the exam, the key skill is knowing when the Agent API fits and when another tool does.

Three jobs: build, test and use

Agentforce developer tools do three jobs: they build agents, test agents or use agents. Using an agent means talking to it. The Agent API is a use tool, because it lets you chat with agents through a REST API. It does not build or test them.