Skip to content

Core Concepts

Understanding AG2Trust's core concepts will help you design effective agent architectures.

Overview

AG2Trust is built around these key concepts:

graph TD
    O[Organization] --> P[Providers]
    O --> AT[Agent Types]
    O --> T[Teams]
    AT --> A[Agents]
    T --> D[Deployments]
    D --> A
    A --> W[Workflows]

Key Concepts

  • Agents

    Individual AI agents powered by LLMs that process messages and perform tasks

    Learn more

  • Teams

    Groups of agents that can collaborate and communicate with each other

    Learn more

  • Agent Types

    Templates defining agent configurations, capabilities, and API endpoints

    Learn more

  • Providers

    LLM providers (OpenAI, Anthropic) that power your agents

    Learn more

  • Workflows

    Automated multi-step processes orchestrating agent interactions

    Learn more

  • Security Model

    RBAC, encryption, and compliance features for enterprise deployments

    Learn more

Resource Hierarchy

Understanding resource ownership is crucial:

Concept Owned By Description
Organization - Top-level entity (your company)
User Organization Individual team member
Provider Organization LLM provider configuration
Agent Type Organization Agent template/blueprint
Team Organization Agent grouping
Agent Organization Running agent instance
Workflow Organization Automated process

Organization-Owned Resources

All resources belong to the Organization, not individual users. When a user creates an agent, the organization owns it. This ensures continuity when team members change.

Quick Reference

Creating Resources

Resource Location Typical Flow
Provider Settings > Providers Add API key for OpenAI/Anthropic
Agent Type Agent Types page Define template, set endpoint slug
Team Teams page Create team, add deployments
Agent Agents page or Team Create standalone or via deployment
Workflow Workflows page Design multi-step automation

Common Patterns

Pattern 1: Standalone Agent 

Provider → Agent Type → Agent (standalone)
Best for: Single-purpose bots, testing

Pattern 2: Load-Balanced Pool 

Provider → Agent Type (with endpoint) → Team → Deployment (replicas: 3)
Best for: Production APIs with traffic

Pattern 3: Collaborative Team 

Provider → Multiple Agent Types → Team → Multiple Deployments
Best for: Complex workflows requiring agent collaboration