Got AI questions? This little glossary just might have some answers! Here's your chance to stay on top of AI terms and see how these concepts integrate with your workflows...
- agentic AI coding tool
A type of AI agent built for software development, such as Claude Code or Cursor. With UE and UEFN, you can connect an AI agent coding tool using MCP. These are sometimes called AI agents.
- agentic workflow
An automated process where an AI agent works toward a task you define. The agent plans the steps, uses tools to carry them out, and checks and corrects its own work, all within the limits you set.
- AI agent
An AI agent is a software system that can operate independently of human instruction or control once you've provided a prompt by breaking down a user's goal into a structured plan of action, running an external tool that acts on that plan, capturing the output, then adjusting, evaluating, and, if indicated, repeating the steps.
- artificial intelligence
1. Artificial intelligence, or AI. In Fortnite, an AI is a programmed opponent, ally or other type of non-player character.
2. More broadly, AI is the ability of a computer system or systems to learn, solve problems, perceive information, and make decisions without the direct intervention of a human agent. It involves creating machines and software capable of simulating human cognitive functions, analyzing data, recognizing patterns, and adapting to new information to achieve specific goals.
3. With the advent of AI agents and the agentic workflow in Epic's game development engines, AI can also refer to tool sets used to automate areas of game development by using prompts that integrate with the editor UI.
- deterministic
A deterministic system is one where the same initial state or input will always produce the same results.
Compare to non-deterministic.
- large language model (LLM)
A type of AI model trained on large amounts of text. It can generate a text output, such as code or instructions, based on a prompt that you input.
- LLM-driven NPC (gameplay)
An NPC whose dialogue or behavior is generated from an LLM at runtime instead of from scripted logic.
- MCP tool
The MCP tool is a single editor action, such as Exec Python, that an AI agent can call through the UEFN MCP.
A toolset is a group of related MCP tools.
- model context protocol (MCP)
Model context protocol, or MCP, is an open standard for connecting AI applications to external tools and data. You can use MCP to connect an AI agent like Claude or Codex to your editor, then use the agent to generate elements of your world using prompts.
- non-deterministic
A deterministic system produces a predictable result each time, but a non-deterministic system, such as a large-language model, can produce random results, even when the same prompts are used. Compare to deterministic.
- NPC behavior (gameplay)
Systems that control how non-player characters (NPCs), such as enemies, wildlife and bots, perceive, move, and act. Examples are behavior trees, StateTree and smart objects.
These systems are separate from and should not be confused with MCP and AI agents.
- prompt
An AI prompt is an instruction, question, or other input that you give to an AI system, such as a chatbot or large language model (LLM), that generates a response or completes a task. The more concise your prompt, the better your result.
- UEFN model context protocol
The MCP server built into UEFN. It makes editor actions available as MCP tools that an AI agent can call.