Zero-Install & Cloud Paths
Last reviewed
BeginnerWhat you'll learn
~10 min- Tell no-repo browser builders apart from repo-backed cloud agents and Linux devboxes
- Understand what each option can and cannot do
- Choose the right cloud sandbox based on your constraints
- Know when an agent on your own computer unlocks more power
Already installed an agent on your computer in Lesson 1.2? You can skim or skip Lessons 3.4 and 3.5; they are here for locked-down work laptops, Chromebooks, and zero-install setups. If your sandbox gives you a terminal, Module 4 teaches the commands you need.
Can’t install software, or don’t want to set anything up yet? You can still build real things with AI tools: here’s how to choose.
- No GitHub account, and you want something quick to share? Start with a no-repo browser builder: Claude Artifacts on claude.ai or ChatGPT Work & Sites on chatgpt.com.
- Your project is (or can be) on GitHub? Claude Code on the web (
claude.ai/code) gives you the closest thing to the full agent with nothing to install; it needs a paid Claude plan (Pro, Max, Ultra, or Team). GitHub Codespaces gives you a whole Linux machine with a free monthly allowance. - You can install apps, you just want to avoid the terminal? You may not need this lesson: a desktop agent app (Lesson 1.2, and Module 6 in depth) works on folders on your own computer with no terminal.
Why this lesson exists
The rest of this course teaches workspace agents that run on your own computer, as a desktop app or a terminal CLI. But not everyone can install software: locked-down work computers, Chromebooks, school labs, and mobile devices all make local installation difficult or impossible.
The good news: as of November 2026 there are two very different kinds of zero-install option, and it helps to keep them apart:
| No-repo browser builders | Repo-backed cloud agents & Linux devboxes | |
|---|---|---|
| Examples | ChatGPT Work & Sites, Claude Artifacts, Lovable, Bolt, v0, Replit Agent | Claude Code on the web, Codex Cloud, Copilot in the cloud, Jules, GitHub Codespaces, Google Cloud Shell |
| What you need | An account with the builder | A GitHub project (for the cloud agents and Codespaces) or a Google account (Cloud Shell) |
| What you get | A live page, dashboard, or small app hosted by the builder, shareable by link | Real project files under version control, changed by an agent or by you in a cloud terminal |
| Best for | Quick prototypes and shareable tools | Real multi-file projects, and practising the same workflow this course teaches |
This lesson helps you choose. The next lesson walks you through setup.
Group 1: No-repo browser builders
These need no GitHub account and nothing installed. You describe what you want, the builder makes it in the browser, and you share a link. Most can export the code or connect a GitHub project later. The tradeoff: the result lives in the builder’s cloud, not in a folder on your computer.
- ChatGPT Work & Sites (chatgpt.com): switch the conversation from Chat to Work, and ChatGPT can build and host a web app, tracker, or dashboard from the spreadsheets and documents you upload.
- Claude Artifacts (claude.ai): Claude builds an interactive page or small app beside the conversation; ask for changes to update it, and share it by link.
- Lovable, Bolt, v0, and Replit Agent: browser builders that turn a description into a full web app with a live preview and one-click hosting.
Best for: A quick shareable page, form, tracker, or dashboard, especially without a GitHub account.
Limitations: Not your local files, and less control over how the app is built and hosted. Plans and limits vary; see the cheat sheet for current pricing.
Group 2: Repo-backed cloud agents & Linux devboxes
These work on a real project with real files, usually kept on GitHub. The cloud agents are the same workspace agents this course teaches, running on a cloud machine against your GitHub project; the devboxes are whole Linux machines in your browser, where you can install and run the terminal CLIs yourself.
Claude Code on the web
What it is: Anthropic’s Claude Code, running in your browser at claude.ai/code. It connects to your GitHub repositories and can clone, edit, test, and commit code — all without a terminal.
Best for: Full project editing via GitHub. The closest browser equivalent to the desktop app and terminal CLI.
Limitations: Built for GitHub repos (other git hosts need an advanced terminal upload, and results can’t be pushed back), requires a paid plan, no offline access. See the cheat sheet for current pricing.
Codex Cloud
What it is: OpenAI’s Codex agent running at chatgpt.com/codex. It plans, executes, and iterates on coding tasks in a sandboxed cloud environment connected to your GitHub repository.
Best for: Handing off a well-defined task and reviewing the result.
Limitations: Each task gets its own workspace, so one task’s changes don’t carry into the next, though a saved environment can reuse your repositories and dependencies. Cloud tasks need a paid ChatGPT plan (Plus or higher). See the cheat sheet for current pricing.
Copilot in the cloud and Jules
What they are: GitHub’s Copilot cloud agent (start a task at github.com/copilot or assign it an issue) and Google’s Jules (jules.google) both take a task on a GitHub repository, work on it on a cloud machine, and hand back the changes for you to review as a pull request.
Limitations: Both need a GitHub project. The Copilot cloud agent is on paid Copilot plans only; Jules has a free daily task allowance. See the cheat sheet for current limits.
GitHub Codespaces
What it is: A full Linux development environment running in your browser. You get VS Code, a terminal, file access, and the ability to install any tool — including Claude Code, Antigravity, Codex CLI, and Copilot CLI.
Best for: The most flexible option. Everything a local terminal can do, in the browser.
Limitations: Free tier has limited hours, sessions time out after inactivity, codespaces auto-delete after extended inactivity. See the cheat sheet for current limits.
For detailed setup instructions, see Lesson 4.5: Remote Environments & VS Code. This lesson covers the decision; that lesson covers the hands-on setup.
Google Cloud Shell
What it is: A free browser-based Linux terminal at shell.cloud.google.com, pre-installed with common dev tools. It needs only a Google account, not GitHub.
Best for: The fastest zero-setup terminal. Great for Module 4 command practice.
Limitations: Idle timeout, weekly hour limit, limited for full project development. Native CLI installers write into your persistent home directory (~/) so tools survive session restarts. See the cheat sheet for current limits and Lesson 4.5 for more details.
Also useful: upload-and-chat
Two chat tools handle quick jobs on a single file without any project at all:
- ChatGPT data analysis runs Python on files you upload and hands you the results to download. Great for single-file tasks, data analysis, and file conversion; no persistent file tree, and no shell where you could install CLI tools or run
git. - Gemini (
gemini.google.com) has a free tier for conversational tasks, brainstorming, planning, and research. It cannot run shell commands, keep a file tree, or install packages.
For current pricing, quotas, persistence details, and capability comparisons, see the Cloud Sandbox Cheat Sheet. That page is the single reference for volatile specs.
Which sandbox fits your situation?
Scenario 1: Locked-down machine (work, school, or government)
You can’t install software, and your IT department controls your environment.
Best option: Codespaces. It runs entirely in the browser, gives you a full Linux terminal, and lets you install CLI tools inside the codespace. If GitHub is blocked, the repo-backed cloud agents won’t help either, because they clone from and push to GitHub; use Cloud Shell for terminal practice and a no-repo browser builder (if your organization allows it) for quick shareable tools.
Scenario 2: Personal machine, but you want to avoid the terminal
You could install tools, but you’d rather not type commands yet.
Best option: a desktop agent app (Lesson 1.2, and Module 6 in depth): install it like any other app, open a project folder, and you have the full agent on your own files with no terminal. If you’d rather install nothing at all, Claude Code on the web gives you the agent on a GitHub project, and a no-repo browser builder is fastest for something quick to share.
Scenario 3: You just want something quick to share
A sign-up form for an event, a small tracker, a dashboard from a spreadsheet, when you don’t need the files on your own computer.
Best option: a no-repo browser builder: Claude Artifacts or ChatGPT Work & Sites for the simplest path, or Lovable, Bolt, v0, or Replit Agent for a fuller web app with hosting.
Scenario 4: You want full terminal power from the browser
You want the complete CLI experience — multiple tools, shell commands, package installation — without any local setup.
Best option: Codespaces. It is a real Linux machine. Install Claude Code, Antigravity, Codex CLI, and Copilot CLI just like you would locally. As of November 2026, a free GitHub account includes 120 core-hours of compute a month (60 hours on a two-core machine) (GitHub’s Codespaces billing page), which covers regular learning sessions.
🏛️In Your Field: Government Devclick to expand
Data sovereignty warning: All cloud sandboxes process your code on external servers. Do not paste sensitive, PII, CUI, or classified data into public sandboxes without authorization from your security office. Government machines often block software installs and may restrict certain web services. Check whether your agency allows access to claude.ai, github.com, or chatgpt.com. If GitHub is accessible, Codespaces may be your best path — it runs entirely in the browser and does not require local installation. Talk to your IT team about approved cloud development environments. A sanctioned environment is not a CUI determination: CUI must not go into AI systems that are internet-enabled or external to agency control (ISOO Notice 2026-01). See approved tools and endpoints.
📊In Your Field: MIS / Businessclick to expand
Many university computer labs restrict software installation. Codespaces is ideal for class projects: every student gets the same environment and no setup conflicts. Verified students get free Codespaces use up to 180 core hours a month (GitHub Classroom docs), and if your professor sets Codespaces as the editor for a GitHub Classroom assignment, that usage is billed to the classroom, not to you.
🧬In Your Field: Biotechclick to expand
If your lab uses shared workstations where you cannot install software, Codespaces gives you a clean environment for bioinformatics scripting. For quick data processing tasks (parsing CSVs, reformatting FASTA files), ChatGPT data analysis works well: upload the file, describe the transformation, download the result.
Honest tradeoffs
What you gain with zero-install options:
- Zero setup time — start building immediately
- Works on any device with a modern browser
- No risk of breaking your local environment
- Easy to share your environment with collaborators (Codespaces)
What you lose:
- Local file access (your files live in the cloud, not on your machine)
- Offline capability
- Full control over your development environment
- Some automation features that depend on local shell access
For most tasks in this course, Codespaces and the repo-backed cloud agents work well, because they run the same agents on real project files. No-repo builders are great for quick tools but don’t teach the project workflow. An agent on your own computer (desktop app or terminal CLI) becomes important when you need your local files, offline access, deep local integration, or maximum performance.
🔍Devcontainer templates: pre-configured sandboxesclick to expand
If you’re using Codespaces, devcontainer templates can pre-configure your entire environment (tools, extensions, settings) in one click. Community-maintained templates can set up several AI CLI tools automatically; check that a template was updated recently and installs the tools this course uses (Gemini CLI, for example, no longer serves personal accounts). This is an advanced option; the basic Codespaces setup from this lesson is all you need to get started. Explore devcontainer templates later when you want a reproducible, shareable environment.
When you’re ready for more
An agent on your own computer unlocks more flexibility: local file access, offline work, custom automation, and the full power of every tool. If you started here because you could not install software, that is completely fine. When your situation changes, Lesson 1.2 installs a desktop agent app or a terminal CLI, Module 6: Your Agent In Depth goes further, and Module 4: Command Line Basics teaches the few terminal commands you need for the CLI.
The orchestration skills you learn in this course (planning, prompting, reviewing, iterating) transfer to any tool, in the browser, in a desktop app, or in a terminal. Start where you can, grow when you are ready.
Key takeaways
- Two kinds of zero-install option: no-repo browser builders (ChatGPT Work & Sites, Claude Artifacts, Lovable, Bolt, v0, Replit Agent) and repo-backed cloud agents & Linux devboxes (Claude Code on the web, Codex Cloud, Copilot in the cloud, Jules, Codespaces, Cloud Shell)
- No GitHub? A no-repo builder gets you a shareable page or app fastest
- Claude Code on the web is the closest browser equivalent to the desktop app and CLI: it connects to your GitHub repos and edits code directly
- Codespaces is the most flexible — it is a real Linux machine where you can install any CLI tool from this course
- Cloud Shell is the fastest zero-setup terminal — great for practicing commands
- An agent on your own computer is still the most powerful. When you can install software, Lesson 1.2 sets up a desktop agent app or a terminal CLI
- Your orchestration skills transfer. Good prompts, clear planning, and thoughtful review work the same way in any tool
Ready to set up your sandbox? Continue to the next lesson to get hands-on.