Egor Urvanov

Terminal

Agents and tools · command line, CLI

A window where commands are typed as text, also called the CLI (command-line interface). An AI CLI agent works right in the terminal: it sees the same files and programs you do and runs them itself — tests, builds, git. Access is limited by permission rules.

Terminal — ~/shop~/shop $ claudeyou› fix the failing test● reading src/checkout.tsagent$ npm test✗ 1 failed: total is NaN● editing src/checkout.ts$ npm test✓ 12 passedAllow "git push"? (y/n)permissions● model, people, systems, data  ● agent and harness  ● risk, stop, boundaryThe agent types commands itself, with the same files as you

AI CLI agents

  • Claude Code · Anthropic’s agent: you start it with the claude command in the project folder; on a Claude plan or pay-as-you-go API.
  • Codex CLI · OpenAI’s agent: the codex command, included in ChatGPT plans.
  • OpenCode · a free open-source agent: works with models from many vendors and with local ones; you pay only for the model.
  • Gemini CLI · Google’s open-source agent on Gemini models; since June 18, 2026 it works only for companies and paid API keys.
  • Antigravity CLI · Google’s agent; since June 18, 2026 it replaced Gemini CLI for individual users, and it is closed-source itself.
  • Hermes Agent · Nous Research’s free open-source agent: runs on your server with any model; you control it from the terminal or through messengers.
  • How it differs from an agent in an IDE · no editor needed: you review edits in your own editor or in git, and you can run the agent from scripts.

Source: Wikipedia: Command-line interface

Where it appears · 2

Related terms

Whole glossary

© 2026 Egor Urvanov