Terminal and CLI tools

Terminal tools bring AI to the place many developers already spend their day. Some are full terminals with AI built in, such as Warp, which can suggest a command from a description or explain an error message. Others are command-line agents and pair programmers, such as Gemini CLI and Aider, that read and edit files in the current repository. Smaller tools such as LLM let you call a model from a script or a pipe like any other command.

The appeal is fit with existing habits. Git, build tools, test runners and package managers are all right there, so a terminal tool can run them directly and you can inspect every step in the scrollback. Aider, for example, commits each change it makes, so undoing a step is an ordinary git command.

The risk is the same thing seen from the other side: a tool that runs shell commands can do real damage. Keep confirmation prompts on for destructive commands, avoid running these tools in a shell that holds production credentials, and read what a suggested command does before you press enter. For scripted use, log prompts and outputs so you can see later what happened.

4 AI coding tools in terminal and cli tools

Warp

Modern terminal with built-in AI that suggests commands, explains errors and runs agent tasks in your shell.

  • Free plan or free version

Gemini CLI

by Google

Google's open-source command-line agent that brings Gemini models into the terminal for coding and other tasks.

  • Free plan or free version
  • Open source

Aider

Open-source AI pair programmer for the terminal that edits files in your git repository and commits each change.

  • Free plan or free version
  • Open source

LLM

by Simon Willison

Open-source command-line tool and Python library for running prompts against hosted and local models and logging the results.

  • Free plan or free version
  • Open source

Pick a terminal tool by how much control you want. A terminal with AI suggestions keeps you in charge of every command. A command-line agent can carry out a whole change while you watch. A scripting tool such as LLM sits inside your own automation, for example to summarize logs or draft commit messages in a pipeline.

Model choice matters here too. Several of these tools let you choose between hosted models and local ones, and some charge nothing themselves while you pay the model provider directly. Check where prompts go, how keys are stored on your machine, and whether the tool keeps a history you would rather not leave on a shared server.

Questions about terminal and cli tools

Can a terminal AI tool run commands without asking?

Many can if you allow it, and most ask for confirmation by default. Leave confirmation on for anything that changes files outside the project, touches remote systems or deletes data.

Which terminal tools are open source?

Gemini CLI, Aider and LLM are open source. Warp is a commercial terminal with a free plan. Open-source tools let you read exactly what they send to a model.

Do terminal tools work with local models?

Several do. Aider and LLM can talk to local models served through tools such as Ollama, which keeps prompts and code on your machine.

Is the terminal or the editor a better place for AI help?

It is mostly preference. The terminal suits people who script their workflow and live in git. Editors suit visual review of diffs. Many developers use both for different jobs.

Related categories