> ## Documentation Index
> Fetch the complete documentation index at: https://docs.refold.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Chains
> Predefined API sequences with conditional logic that provide dynamic context for intelligent agent execution
**Smart Sequencing**: Chains guide agents through logical API call sequences based on business context.
## Setup
### Adding a Chain to Your MCP Server
Go to your MCP server configuration where you have already added APIs and Workflows.
Click "Add Chain" to create a new business process sequence.
Define your chain configuration with business goals and execution logic.
### Chain Configuration Fields
**Chain Name**
Provide a clear, descriptive name for your chain that reflects the business process.
**Examples:**
* "Salesforce Lead Qualification Process"
* "Purchase Order Creation and Approval"
* "Customer Onboarding Workflow"
**Business Objectives**
Define what business goal this chain achieves for the agent.
**Examples:**
* "Qualify inbound leads and assign to appropriate sales rep"
* "Create purchase orders with proper approvals and vendor validation"
* "Onboard new customers across CRM and billing systems"
**API Sequence & Conditions**
Define the sequence of APIs to call with any conditional logic.
**Example Structure:**
* Step 1: Check if contact exists in CRM
* Step 2: If contact missing, create new contact
* Step 3: Create opportunity with contact validation
* Step 4: Assign to sales rep based on territory rules
**Prerequisites Validation**
Define any validations to check before running this sequence.
**Examples:**
* Verify user has permission to create opportunities
* Check if required fields are provided
* Validate data format and constraints
**Success Actions**
Define what to do when the chain executes successfully.
**Examples:**
* Return opportunity ID and details
* Send notification to sales rep
* Log successful execution for audit
**Failure Recovery**
Define what to do when the chain fails or encounters errors.
**Examples:**
* Retry with exponential backoff
* Rollback partial changes
* Escalate to manual review
* Return detailed error information
## Chain vs Individual APIs
**Business Process Sequences**
* Predefined logical API sequences
* Include conditional logic and context
* Guide agents through complex workflows
* Share context across all API calls
* Handle business rule validation
**Best for**: Complex business processes requiring multiple steps
**Atomic Operations**
* Single-purpose API endpoints
* Direct system interactions
* Simple input/output operations
* Independent execution context
* Immediate response handling
**Best for**: Simple, standalone operations
## Agent Decision Process
Agent analyzes user request:
* Searches available chains first
* Matches intent with chain capabilities
* Evaluates chain context requirements
Agent selects execution method:
* **Chain found**: Execute predefined sequence
* **No chain match**: Use individual APIs
* **Partial match**: Execute relevant chain portions
Agent follows chain guidance:
* Uses shared context for decisions
* Follows conditional logic paths
* Handles errors with retry mechanisms
**Result**: Chains act as dynamic prompts that guide agents through complex business processes while maintaining flexibility for simple operations.