AI Agent
Also known as: Autonomous agent, LLM agent, Tool-using agent
An LLM-based system that can plan, use tools and take multi-step actions toward a goal — not just answer a single prompt.
An AI agent is a system that wraps a large language model with a planning loop, a set of tools (search, code execution, file editing, web browsing, APIs) and memory. Instead of answering one prompt, the agent decomposes a goal into sub-tasks, calls tools, observes the result, and iterates until the goal is met.
The simplest agents are ReAct-style chat loops with function calling. More advanced agents use task graphs, verification loops, sub-agents, and persistent memory. Examples include coding agents (Claude Code, Cursor, Cline), research agents (Perplexity, Manus) and browser agents (OpenAI Operator, Browser Use).
Reliable agents depend on three things: a strong reasoning model (today usually Claude Sonnet or GPT-5), a well-defined tool surface, and guardrails — sandboxing, per-action approval, and clear stopping conditions.
See also on SoftPerceptron
Related terms
- Large Language Model (LLM)
A neural network trained on massive text corpora to predict the next token, used for chat, coding, reasoning and as the brain inside AI agents.
- Agentic AI
AI systems that act autonomously over many steps to achieve a goal, combining planning, tool use and memory.
- Model Context Protocol (MCP)
An open standard from Anthropic for connecting LLMs to external tools and data sources through a uniform server interface.
- Prompt Engineering
The practice of designing inputs to LLMs to reliably produce useful outputs — through structure, examples, role-setting and constraints.