Learn about AI
Try one clear question in an assistant. Compare the answer with a trusted source. Learn what a model, app, and agent each mean before adding integrations.
Understand the concepts →Start Here
You do not need
every tool.
Choose one path based on what you want to do. You can stop after that path and return when you need another piece.
Choose a path
What are you trying
to do?
Every path starts small. Accounts, paid plans, and APIs are optional until a real need calls for them.
Improve an assistant
Start with a better prompt and a simple example of the output you want. If the same workflow repeats, consider a reviewed skill. Add an API or MCP connection only for a specific task.
Browse the catalog →Build a website
Describe the page, make a small version in a coding tool, save it with Git, preview and test it, then publish through an approved host.
See the workflow →Three calm steps
From “maybe” to
a safe first try.
Name the outcome
Write down what you want to make, learn, or automate. A clear outcome prevents tool-shopping from becoming the project.
Read the trust labels
Check source, last-reviewed date, prerequisites, permissions, cost, and whether an item is link-only or download-eligible.
Try the smallest slice
Start with a reversible example. Do not add credentials, private data, payments, or broad permissions until you understand the workflow.
Plain-language guide
The pieces fit
different jobs.
These are starting definitions, not a ranking. Product features and availability can change; each link goes to a primary source checked on 2026-09-30.
ChatGPT and Codex
ChatGPT is an assistant for questions and work. Codex focuses on software development and technical work.
- When useful
- Start with a question or idea; use a coding tool when the work involves a codebase.
- Nearest comparison
- An assistant helps with a task; a model is the technology an app uses to generate responses.
- Simple example
- Ask ChatGPT to clarify a website idea, then use Codex to edit and test its code.
Claude and Claude Code
Claude is Anthropic’s conversational assistant. Claude Code is its tool for working with code and developer workflows.
- When useful
- Use the chat product for discussion; use Claude Code when the task needs code or project files.
- Nearest comparison
- The chat product and the coding tool have overlapping ideas but different working surfaces.
- Simple example
- Discuss a feature in Claude, then review a repository change in Claude Code.
Cursor
Cursor is an AI-powered code editor that works with project files.
- When useful
- Choose it when you want to read and change code in an editor.
- Nearest comparison
- An editor is a place to work on files; Git records changes to those files.
- Simple example
- Open a website project in Cursor and ask for a small page change.
Git and GitHub
Git tracks versions of files. GitHub hosts Git repositories and supports sharing and review.
- When useful
- Save a working version before a change; use a hosted repository to collaborate.
- Nearest comparison
- A repository holds source and history. A deployed site is the running result people visit.
- Simple example
- Commit a page change with Git, then review it on GitHub.
Vercel
Vercel is a platform for building and deploying websites and applications.
- When useful
- Preview a website, then publish an approved version at a domain.
- Nearest comparison
- A host serves the site; a repository stores the code used to build it.
- Simple example
- Deploy a tested website from a repository and check its preview before release.
API and MCP
An API lets software request another system’s capabilities. MCP is a standard way for AI applications to connect to outside data and tools.
- When useful
- Add a connection only when a task really needs outside information or action.
- Nearest comparison
- An API is a service interface; an MCP server can expose data or tools to a compatible AI client. Neither grants unrestricted permission by itself.
- Simple example
- A weather app may call a weather API; an assistant may use an approved MCP weather tool.
Prompt, skill, and plugin
A prompt is an instruction for one interaction. A skill packages reusable guidance and sometimes supporting files. A plugin adds capabilities to a host product.
- When useful
- Begin with a prompt; package repeated, reviewed work only when reuse helps.
- Nearest comparison
- A reusable instruction is not an app, and installing it does not grant permission to act.
- Simple example
- Save a good review prompt; later turn its repeated steps into a documented skill.
One small example
From idea to website.
- Describe a useful page and who it helps in an assistant.
- Use a coding tool to make one working page.
- Save the source and changes in a Git repository.
- Preview on a phone and computer; test the links and controls.
- Publish the approved version through a hosting provider.
A domain, API, database, and paid plan are separate choices. Add them when the project actually needs them.
Project guidance
Files that guide
coding tools.
design.md
A project-defined design brief can explain layout, colors, typography, and visual constraints. Its meaning depends on the project that owns it.
AGENTS.md
A repository instruction file supported by coding agents. It can describe how to work in that project; it does not create new permissions.
Official guide ↗CLAUDE.md
Claude Code’s project context file. Exact loading rules depend on Claude Code and its configuration.
Official guide ↗Sources checked 2026-09-30. Instructions and downloaded skills still need review; they do not grant permission to access data or take actions.
Know the labels
Review state is part
of the resource.
- Approved public
- Reviewed for this listing and approved for public display. Check the source for current details.
- Link only
- Use the canonical source. AroraIndex does not redistribute the artifact.
- Download eligible
- Rights and package checks permit inclusion in an AroraIndex bundle.
Ready?