A coding agent from Anthropic that runs in your terminal: it reads your project, edits files and runs commands with your permission.

OpenAI's coding agent. It has an open-source CLI and versions for the editor and the cloud; it takes a task and does it on your code.

A code editor from Google designed to work with agents, which plan, write and test while you supervise.

An open-source coding agent for the terminal that works with different model providers, including local models.

Do I need to know how to code to use an agent with a repo?

It helps, above all to understand what the agent proposes. You can ask it to explain each step before it runs it.

Is it safe to let an agent install a repo?

It can be if you take precautions: work on a copy or in an isolated environment, do not paste keys or passwords into it and review what it is going to run.

Home / Tools

Claude Code, Codex, Antigravity, OpenCode and GitHub, no fluff.

What each one is, what it is for and what to watch out for. These are the tools that come up again and again on the channel when it comes to reading, installing and trying repos.

By Jorge Aguilar · Summary based on official documentation · Updated October 29, 2026

Start with Claude Code Checklist for checking a repo

The five tools

Claude Code

Anthropic

Official repo on GitHub

What it is: A coding agent that lives in your terminal. It reads your project, proposes changes, edits files and runs commands, asking your permission on the important steps.
What it is for: Understanding a repo you did not write, installing it and checking that what the README says is true.
What to watch: Work on a branch or a copy and review every change before accepting it.

Codex

OpenAI

Official repo on GitHub

What it is: OpenAI's coding agent. There is an open-source CLI for the terminal and versions for the editor and the cloud: you give it a task, it does it on your code and shows you the changes.
What it is for: Delegating concrete tasks and comparing the result with another agent's.
What to watch: Define the task and the files it may touch clearly; a vague instruction gives vague changes.

Antigravity

Google

Official site

What it is: A code editor from Google designed to work with agents: the agent plans, writes and tests while you supervise from the editor itself.
What it is for: Working with agents without leaving a full editor, with their plan and their results in view.
What to watch: It is new and changes fast: check usage limits and permissions before giving it access to your project.

OpenCode

Open source

Official site

What it is: An open-source coding agent for the terminal. It works with different model providers, including local models, so it does not tie you to a single company.
What it is for: Choosing the model yourself, or using a local one when you do not want to send your code to the cloud.
What to watch: If you use a local model, check how much memory it needs and what license that model has.

GitHub

Go to GitHub

What it is: The platform where almost all repos live. There you see the code, the license, the history, the issues and the published releases.
What it is for: Finding a repo, reading its README and deciding whether it is worth it before downloading anything.
What to watch: Stars measure fame, not quality: look at the license, the latest activity and whether it is archived.

Which one to pick, depending on what you want to do.

I want…	Start with
An agent in the terminal that reads my project and makes changes with my permission	Claude Code, Codex or OpenCode
A full editor built to work with agents	Antigravity
To choose the model, or use a local one	OpenCode
To find a repo, read its license and download it	GitHub
To understand a new repo without reading it line by line	Any agent, asking it to explain the repo before touching anything

Products change fast: confirm features and limits on each one's official site.

Before you let an agent touch your code.

Work on a branch or a copy.That way you can throw everything away with one command if something goes wrong.
Ask it to explain before it runs anything.“Tell me what this script does and what it will install” avoids surprises.
Review every change.Read the full diff before accepting it; compiling does not mean it is right.
Do not paste keys, tokens or passwords into it.Use environment variables and files the agent does not have to read.
Test the important parts yourself.Run the tests and check by hand anything that touches money, data or security.

See the “Code with AI” series (videos in Spanish)

Frequently asked questions.

What is Claude Code?A coding agent from Anthropic that runs in your terminal: it reads your project, edits files and runs commands with your permission.
What is Codex?OpenAI's coding agent. It has an open-source CLI and versions for the editor and the cloud; it takes a task and does it on your code.
What is Antigravity?A code editor from Google designed to work with agents, which plan, write and test while you supervise.
What is OpenCode?An open-source coding agent for the terminal that works with different model providers, including local models.
Do I need to know how to code to use an agent with a repo?It helps, above all to understand what the agent proposes. You can ask it to explain each step before it runs it.
Is it safe to let an agent install a repo?It can be if you take precautions: work on a copy or in an isolated environment, do not paste keys or passwords into it and review what it is going to run. Checklist for checking a repo.

Claude and Claude Code are trademarks of Anthropic; Codex and OpenAI, of OpenAI; Antigravity, of Google; OpenCode, of its authors; GitHub, of GitHub, Inc. IA de Guardia is independent and is not affiliated with or sponsored by any of them. Updated October 29, 2026.