Flow Hierarchical Record Creator

A generic Salesforce Flow Invocable Action that creates up to 3 levels of related records (parent → child → grandchild) in a single atomic transaction — with automatic lookup field wiring between levels.

By Redpoint Ascent

The Problem

Screen Flows defer DML until a commit point (a Screen element or flow end). This means when you create parent, child, and grandchild records in sequence, the record IDs aren't available to populate lookup fields on subsequent records. Your child and grandchild records get created without their lookup relationships.

Common workarounds either break atomicity (adding screens between creates) or rely on Aura components (UnofficialSF's CommitTransaction) that don't work in LWR Digital Experience Sites.

The Solution

This invocable action accepts up to 3 collections of sObject records, inserts them in order, and automatically populates the lookup fields between levels using the record IDs generated from each insert. If anything fails at any level, the entire transaction rolls back — no orphaned records.

Key Features

• Generic — Works with any standard or custom objects. No hardcoded references.
• LWR Compatible — Pure Apex. No Aura or LWC dependencies.
• Atomic — All-or-nothing. Full rollback on any failure.
• Flow-native errors — Failures surface on the Flow's Fault connector.
• Bulk-safe — Each level is inserted as a single bulk DML operation.

Compatibility

Environment	Supported
LWR Digital Experience Site	✅
Aura Digital Experience Site	✅
Lightning App (internal)	✅
Screen Flow	✅
Autolaunched Flow	✅
Record-Triggered Flow	✅
Salesforce Mobile App	✅

Install

Option 1: SFDX Source Deploy

git clone https://github.com/redpointascent/flow-hierarchical-record-creator.git
cd flow-hierarchical-record-creator
sf project deploy start --source-dir force-app --target-org your-org-alias

Option 2: Metadata API / Workbench

Deploy the contents of force-app/main/default/classes/ to your org using Workbench, VS Code, or your preferred deployment tool.

Option 3: Copy & Paste

Copy HierarchicalRecordCreator.cls and HierarchicalRecordCreatorTest.cls into your org via Developer Console or VS Code. Create the corresponding -meta.xml files with API version 58.0.

Quick Start

1. Deploy the classes to your org.
2. Open your Screen Flow in Flow Builder.
3. Build your parent record collection (e.g., a list of Accounts).
4. Build your child record collection and a parallel index list mapping each child to its parent's position (0-based).
5. Add an Action element → search for "Create Hierarchical Records".
6. Map your inputs.
7. Add a Fault connector to handle errors.

For a detailed walkthrough, see the Usage Guide.

Inputs

Input	Type	Required	Description
parentRecords	List<SObject>	Yes	Level 1 records to insert
childRecords	List<SObject>	No	Level 2 records to insert
childLookupField	String	If children	API name of lookup field on child → parent
childParentIndex	List<Integer>	If children*	0-based index mapping each child to its parent
childMatchField	String	If children*	Field on child records containing the match key
parentMatchField	String	If children*	Field on parent records containing the unique match key
grandchildRecords	List<SObject>	No	Level 3 records to insert
grandchildLookupField	String	If grandchildren	API name of lookup field on grandchild → child
grandchildParentIndex	List<Integer>	If grandchildren*	0-based index mapping each grandchild to its child
grandchildMatchField	String	If grandchildren*	Field on grandchild records containing the match key
grandchildParentMatchField	String	If grandchildren*	Field on child records containing the unique match key for grandchild matching

*At each level, provide either the index map or both match fields — not both. See Field-Based Matching below.

Outputs

Output	Type	Description
parentIds	List<Id>	IDs of inserted parents (same order as input)
childIds	List<Id>	IDs of inserted children (same order as input)
grandchildIds	List<Id>	IDs of inserted grandchildren (same order as input)
isSuccess	Boolean	true if all inserts succeeded
errorMessage	String	Exception details on failure

How the Index Mapping Works

The index list is a parallel collection where each value is the 0-based position of the parent record that the child belongs to.

Example: 2 Accounts, 5 Contacts (3 under Account 0, 2 under Account 1):

parentRecords:    [ Account-A,  Account-B ]
                     index 0     index 1

childRecords:     [ Contact-1, Contact-2, Contact-3, Contact-4, Contact-5 ]
childParentIndex: [     0,         0,         0,         1,         1     ]

Contacts 1–3 get AccountId = Account-A.Id, Contacts 4–5 get AccountId = Account-B.Id.

Field-Based Matching

As an alternative to index arrays, you can match children to parents by field values. Instead of building a parallel index list, you specify which field on the child holds a key and which field on the parent holds the corresponding unique key.

Example: Match Contacts to Accounts by AccountNumber:

parentRecords:    [ Account(AccountNumber='ALPHA-001'), Account(AccountNumber='BETA-002') ]
childRecords:     [ Contact(Department='ALPHA-001'), Contact(Department='BETA-002'), Contact(Department='ALPHA-001') ]
childMatchField:     'Department'
parentMatchField:    'AccountNumber'

Contact 1 and 3 get AccountId = ALPHA-001 Account.Id, Contact 2 gets AccountId = BETA-002 Account.Id.

Rules

• Mutually exclusive with index mode — use one or the other at each level (mixing across levels is fine).
• Parent values must be unique — duplicate match keys on the parent side cause an error.
• Child values must match — every child's match value must exist in the parent's match values.
• Case-insensitive — ABC-001 matches abc-001.
• No nulls or blanks — all match field values must be populated.

When to Use Which

Scenario	Recommended Mode
Records built in nested loops with counters	Index-based (childParentIndex)
Records from a data source with natural keys (IDs, codes, emails)	Field-based (childMatchField + parentMatchField)
Migrating from UnofficialSF "Find Record in Collection" pattern	Field-based

License

MIT

Contributing

Issues and pull requests welcome. Please include test coverage for any new functionality.