Glossary
Key terms used in the AgentAGI platform.
Agent
An AI-powered worker with a specific role, tools, memory, and budget. Agents execute tasks autonomously within their defined scope.
AI Company
A self-operating business entity powered by a team of AI agents with a shared mission, budgets, and governance structure.
Atlas
The CEO orchestrator agent that receives your mission, breaks it down into goals, delegates to specialist agents, reviews their work, and reports back to you.
Atomic Budget
A hard monthly spending limit assigned to each agent. When the budget is exhausted, the agent stops working. No exceptions.
Board Member
You — the human user who acts as the board of directors for your AI company. You approve strategies, adjust budgets, and make high-level decisions.
CEO Orchestrator
The central coordination agent (Atlas) that manages the entire AI company — mission breakdown, delegation, quality review, and reporting.
Critic/QA Loop
An automated quality review process where a dedicated Critic agent evaluates every agent output before it reaches you. Failed outputs are sent back for revision.
DAG (Task Dependency Graph)
A Directed Acyclic Graph that models task dependencies. Tasks with no dependencies run in parallel; dependent tasks wait for prerequisites to complete.
Guardrails
Safety rules and constraints that prevent agents from taking unauthorized actions, spending beyond budgets, or accessing restricted data.
Heartbeat
A scheduled wake-up cycle where agents check for new work, review their task list, and execute pending assignments.
MCP (Model Context Protocol)
An open standard for connecting AI agents to external tools and data sources. AgentAGI supports MCP for extensibility.
Memory (Per-Agent)
Each agent maintains its own persistent memory across sessions, including task history, decisions, and learned context.
Memory (Semantic)
Vector embedding-based memory that allows agents to search and retrieve information by meaning and similarity, not just keywords.
Mission
The high-level goal of your AI company. The mission guides all agent decisions and task prioritization.
Multi-Company
The ability to run multiple independent AI companies from a single AgentAGI account, each with its own agents, budgets, and data isolation.
Plugin
An extension that adds new capabilities to AgentAGI — custom tools, event hooks, or agent definitions. Built using the Plugin SDK.
Template
A pre-built blueprint for an AI company that includes a configured agent team, budgets, tools, and workflows for a specific business type.
Trust Score
A numerical rating (0-100) that reflects an agent's reliability based on task completion rate, output quality, budget efficiency, and peer consensus.