Agent Building Glossary

Master the language of AI agents. Explore key concepts, terminology, and VM0-specific features to accelerate your agent development journey.

29 Terms
4 Categories

Showing all terms 29 terms

A

Core Concepts

Agent

An autonomous AI program that can understand instructions, make decisions, use tools, and execute tasks independently. Unlike traditional scripts, agents can adapt their approach based on context and iterate toward goals.

Related terms

ToolSkillSession
Execution

Artifact

Files, data, or outputs generated by an agent during execution. Artifacts persist across runs and can include code, documents, analysis results, or any other deliverables produced by the agent.

Related terms

Checkpoint
Development

AGENTS.md

A markdown configuration file that defines an agent's behavior, personality, goals, and available tools. The AGENTS.md file serves as the agent's blueprint, specifying how it should think, what it can do, and how it should interact with users and systems.

Related terms

AgentSkill
Execution

Action

A concrete operation or task that an agent performs to achieve its goals. Actions can include calling APIs, reading files, processing data, or interacting with external systems. Each action represents a single step in the agent's execution flow.

Related terms

ToolAgentWorkflow
Infrastructure

Agent Router

A system that intelligently routes agent requests to the most appropriate AI model based on task requirements, cost constraints, and performance needs. The agent router can dynamically select between different LLMs (Claude, GPT, Gemini) or route to specialized models for specific capabilities.

Related terms

AgentLLM (Large Language Model)Orchestration

C

Core Concepts

Context

The information available to an agent during execution, including conversation history, memory, tool outputs, and environmental state. Context helps agents maintain coherence across interactions and make informed decisions.

Related terms

SessionMemoryAgent
Execution

Checkpoint

A snapshot of an agent's state at a specific point in time, capturing its memory, context, and execution progress. Checkpoints enable reproducibility, allowing you to restore, fork, or replay agent runs.

Related terms

SessionArtifact
Development

CLI (Command Line Interface)

A text-based interface for interacting with agents and the VM0 platform. The CLI allows developers to create, deploy, and manage agents through terminal commands, providing powerful scripting and automation capabilities without requiring a graphical interface.

Related terms

CLI AgentAgent
Core Concepts

CLI Agent

An agent specifically designed to be invoked and controlled through command-line interfaces. CLI agents can be integrated into scripts, CI/CD pipelines, and terminal workflows, making them ideal for automation tasks and developer tooling.

Related terms

CLI (Command Line Interface)Agent
Development

Claude Code

Anthropic's official command-line tool for building and deploying AI agents. Claude Code provides a developer-friendly interface for creating agents, managing skills, and integrating with the Claude API. It streamlines the agent development workflow with features like live reloading, debugging tools, and deployment automation.

Related terms

CLI (Command Line Interface)AgentCLI Agent
Development

Codex

A command-line tool for agent development and deployment. Codex provides developers with utilities for building, testing, and managing agents through the terminal. It offers features like code generation, scaffolding, and integration with AI models to streamline the agent development workflow.

Related terms

CLI (Command Line Interface)AgentClaude Code

E

Infrastructure

Environment Variable

A configuration value that is passed to an agent at runtime, allowing you to customize behavior without changing code. Environment variables are commonly used for API endpoints, feature flags, and non-sensitive configuration settings.

Related terms

SecretsRuntime

I

Infrastructure

Image

A packaged container that includes the code, runtime, dependencies, and configuration needed to run an agent. Images provide a consistent and reproducible execution environment, ensuring that agents behave the same way across different deployments.

Related terms

RuntimeSandbox

L

Core Concepts

LLM (Large Language Model)

The underlying AI model that powers agent reasoning and decision-making. Examples include Claude, GPT-4, and Gemini. Different LLMs have different strengths, costs, and capabilities.

Related terms

Agent

M

Core Concepts

Memory

Information that an agent retains across interactions, enabling it to learn from past experiences and maintain context over time. Memory can include facts, preferences, conversation history, and learned patterns.

Related terms

ContextSession
Core Concepts

MCP (Model Context Protocol)

An open protocol developed by Anthropic that enables AI assistants to securely connect to external data sources and tools. MCP provides a standardized way for agents to access resources like databases, APIs, and file systems while maintaining security and context awareness.

Related terms

ToolAgentSkill

O

Infrastructure

Observability

The ability to inspect, monitor, and understand agent behavior through logs, metrics, and execution traces. Observability is essential for debugging, optimizing, and ensuring transparent AI operations.

Related terms

Agent
Development

Orchestration

The coordination and management of multiple agents, workflows, or tasks to achieve complex goals. Orchestration handles dependencies, sequencing, parallel execution, and resource allocation across distributed agent operations.

Related terms

AgentWorkflow

P

Core Concepts

Prompt

Natural language instructions given to an agent to define its task, behavior, or goal. Prompts can range from simple commands to complex system instructions that shape how an agent thinks and operates.

Related terms

AgentContext

R

Infrastructure

Runtime

The execution environment where agents run, providing the necessary infrastructure, resources, and APIs. The runtime manages agent lifecycle, handles resource allocation, and ensures proper isolation between different agent executions.

Related terms

SandboxAgentSession

S

Core Concepts

Skill

A pre-built integration or capability that agents can use to interact with external services. Skills provide agents with ready-made functionality like accessing Slack, GitHub, Notion, or other APIs without requiring custom implementation.

Related terms

ToolAgent
Execution

Session

A continuous period of agent execution that preserves state, memory, and context. Sessions allow agents to maintain awareness across multiple interactions and resume work without losing progress.

Related terms

ContextCheckpoint
Infrastructure

Sandbox

An isolated execution environment where agents run safely without affecting production systems. Sandboxes provide security boundaries and ensure that agent actions are contained and controlled.

Related terms

AgentSession
Development

SKILL.md

A markdown configuration file that defines a skill's behavior, inputs, outputs, and implementation logic. The SKILL.md file allows you to create custom skills using natural language and structured metadata, making it easy to extend agent capabilities without writing complex code.

Related terms

SkillAgent
Infrastructure

Secrets

Sensitive information like API keys, passwords, and tokens that agents need to access external services. Secrets are stored securely and injected into the agent runtime at execution time, preventing exposure in code or configuration files.

Related terms

Environment VariableAgent

T

Core Concepts

Tool

A function or API that an agent can call to perform specific actions. Tools are the building blocks of agent capabilities, enabling them to read files, make HTTP requests, query databases, or interact with any external system.

Related terms

SkillAgent

V

Infrastructure

Volume

A persistent storage space that preserves files and data across agent runs. Volumes provide agents with a file system that persists beyond individual executions, enabling them to store state, cache data, and maintain long-term artifacts.

Related terms

ArtifactSession
Infrastructure

Virtual Machine

An isolated, software-based computing environment that emulates a physical computer. Virtual machines provide complete isolation between agent executions, ensuring security and preventing interference. VM0 uses lightweight virtual machines to run agents with their own operating system, file system, and network stack.

Related terms

SandboxRuntimeImage

W

Development

Workflow

A sequence of tasks or operations that an agent executes to achieve a goal. Unlike rigid automation scripts, agent workflows can adapt and adjust based on intermediate results and changing conditions.

Related terms

Agent
Agent Building Glossary - VM0 | VM0