Glossary
What is agentic AI? A short glossary
Plain-language definitions for terms that come up when people talk about agentic AI. Written for anyone curious — no prior background assumed.
- Agent
- A software system that takes actions towards a goal, often using AI models to decide what to do next.
- Tool use
- When an agent calls external functions, APIs, or scripts to gather information or take actions.
- Context window
- The amount of text or data an AI model can consider at once when producing a response.
- Guardrail
- A rule, filter, or check that constrains what an agent is allowed to do or say.
- Human-in-the-loop
- A workflow where a person reviews or approves an agent's important actions before they take effect.
- Evaluation
- Structured tests that measure how well an agent performs on specific tasks, including where it fails.