Agent Fundamentals & Architecture

Core definitions and Architectural Components

In the Mattr ecosystem, the Agent Builder is the control center for creating your automation's intelligence. An Agent is an intelligent entity designed to execute complex, multi-step tasks that require reasoning, knowledge, and the ability to use specialized tools (Skills).

The Agent architecture mirrors the structure of a human brain, ensuring the entity can reason and act autonomously within your enterprise context:

Component	Function	Detail
Reasoning Engine	The Brain / Core Logic	Processes incoming information, manages the execution sequence, and decides the next course of action.
Knowledge	The Memory / Data Store	The structured information (documents, data) the Agent uses to perform fact-based, contextual tasks.
Skills	The Capabilities / Tools	The reusable abilities the Agent executes (e.g., calling an API, reading a document) to perform actions.
Memory	Short-term & Long-term Context	Retains context during a specific interaction (short-term) and continuously learns and adapts from user feedback and past evaluations (long-term).

Agent Configuration Settings

Before an Agent starts executing, its core personality and operating style must be defined. These configurations set the balance between speed, cost, and accuracy, dictating how the Agent operates within a workflow :

Setting	Purpose	Key Trade-off / Options
Agent Mode	Determines the balance between speed, cost, and reasoning quality.	Flash: Lightning speed, instant responses. Advanced: Maximum accuracy, premium reasoning. Balanced: Cost-effective and steady accuracy.
Autonomous Level	Sets the amount of human intervention required during task execution.	Controls how often the Agent will ask for permission or guidance. (Low, Medium, High).
Reasoning Strategy	Defines the underlying approach the Agent uses to solve a problem.	Intelligent Planner: Agent creates a complete plan first. Next Best Step: Agent determines only the next immediate action based on the current state.