Glossary

Agentic Coding

A workflow where an autonomous AI agent plans and executes multi-step coding tasks, not just single completions.

Agentic coding describes development workflows where an AI agent operates autonomously across multiple steps: reading the codebase, planning changes, editing files, running commands, and checking results — looping until a goal is met. It contrasts with single-shot autocomplete, where the model only suggests the next few lines.

Agentic coding tools (Claude Code, Codex CLI, Gemini CLI, Aider) live in the terminal because that's where building, testing, and shipping happen. The terminal you run them in shapes how much you can orchestrate and observe.

Learn more

Related guides on the Agents UI blog and related terms in this glossary.

Try Agents UI

A native terminal for AI coding agents with persistent sessions, SSH workflows, and built-in editing.

Download View on GitHub

Related

AI Coding Agent

An AI program that autonomously reads, writes, and runs code to complete development tasks.

Multi-Agent Orchestration

Coordinating several AI agents toward one goal by splitting it into tasks across parallel sessions.

Prompt Engineering

Structuring instructions and context so an AI model produces reliable, useful output.

Agentic Coding Developers Guide Autonomous Ai Agents

Read the guide →

Prompt Engineering For Ai Coding Agents

Read the guide →