Business rules

Overview Business Rules are intelligent agent connectors that execute automated logic, validations, and transformations. They enforce business requirements, automate repetitive tasks, and ensure data consistency. They are executed when a process is created or updated for a sub project. Purpose

Automate business logic enforcement
Execute intelligent decision-making
Validate data against business requirements
Transform and enrich data automatically
Support complex conditional processing

Business Rules Architecture Business Rules operate through Agent Connectors, which are groupings of multiple agents that work together to accomplish specific business objectives. Business Rule Execution Triggers

Trigger	Timing	Use Case
On Create	When new record created	Initialize defaults, validate inputs, trigger workflows
On Update	When record modified	Validate changes, cascade updates, trigger reactions

Super Lock An agent could be assigned to apply super lock to a resource for the process. A default behaviour could also be assigned when the agent fails.

Process Context Passing & Consumption

When a business rule (blocking or non-blocking connector) executes, the platform packages the process context into the slots of the agent’s payload. This allows the target agent to access and inspect process details.

Context Structure

The payload sent to the agent is structured in the slots map as follows:

Slot Key	Type	Description
`process`	`object`	The complete serialized process record (containing all fields, stage, status, etc.).
`fieldsConfig`	`object`	The configuration of fields for the sub-project.
`updates`	`object`	The payload of changes being made (available only during `update` triggers).
`metaDetails`	`object`	System metadata for the trigger/request.
`fields`	`object`	Initial field values (available only during `create` triggers).

Consuming Context in the Agent

Within the agent code (e.g., Python code executors), the slots can be accessed directly from metaData. For example:

# Accessing process context inside a user Python agent executor
process_data = metaData.get("slots", {}).get("process", {})
process_fields = process_data.get("fields", {})

# Inspecting modified fields during an update
updates = metaData.get("slots", {}).get("updates", {})

# Validate fields
if process_fields.get("amount") > 10000 and not updates.get("approvedBy"):
    raise Exception("High-value transactions require compliance approval.")

Returning Updates to the Process

To update the process fields from the agent:

Return a dictionary containing a result key (e.g. {"result": {"status": "APPROVED", "remarks": "Auto-verified"}}).
The platform automatically merges this result back into the process’s dynamic fields.
If a blocking agent action fails or raises an error, the platform aborts the process creation/update transaction.

Examples

Validator

Validates data against rules, patterns, or business logic. Blocks action if validation fails.

Transformer

Transforms and enriches data. Calculates computed fields and normalizes values.

External Service

Integrates with external APIs and services. Fetches data and triggers remote actions.

Decision Engine

Makes intelligent decisions based on conditions and business logic. Routes workflows.

Notifier

Sends notifications via email, Slack, webhooks, or other channels. Alerts users and teams.

Lookup

Searches and retrieves data from existing records. Performs duplicate checking.

Overview

Agentic ERP

Agentic Orchestration

Agentic Brain

Resources

Process Context Passing & Consumption

Context Structure

Consuming Context in the Agent

Returning Updates to the Process

Validator

Transformer

External Service

Decision Engine

Notifier

Lookup

​Process Context Passing & Consumption

​Context Structure

​Consuming Context in the Agent

​Returning Updates to the Process

Validator

Transformer

External Service

Decision Engine

Notifier

Lookup

Process Context Passing & Consumption

Context Structure

Consuming Context in the Agent

Returning Updates to the Process