Challenge 3: Install R-CLI (or Backboard Studio) + First Walkthrough

Global Hack Week: Agent Week | Backboard.io

Time needed: about 15 minutes (a bit more on Windows, see below)
Difficulty: Medium. You will use the terminal, but every command is copy-paste.
What you need: Your Backboard account from Challenge 1. macOS, Linux, or Windows with WSL.

WALKTHROUGH VIDEO HERE

The written steps below stand on their own if you would rather read.

---

What is R-CLI?

R-CLI is Backboard’s AI coding agent that lives in your terminal. You type what you want in plain English, and it plans, writes the code, runs it, fixes its own errors, and shows you the result.

The “R” stands for recursive: instead of one AI trying to hold your whole project in its head, R-CLI breaks work into pieces and delegates them to child agents with their own focused context. The harness does the heavy lifting, not just the model. That design is why R-CLI scored 84.3% on Terminal Bench 2.1, the highest publicly reported score on the benchmark, above every published result.

It also ships with hundreds of pre-loaded MCP connections and skills (integrations and playbooks your agent can use out of the box).

For the rest of Agent Week, R-CLI is your power tool. In Challenges 4, 5, and 6 you will not hand-write code. You will describe what you want and let the agent build it. Installing it now is what makes that possible.

Prefer an app over a terminal? See the Backboard Studio section at the bottom. Studio is a desktop app built on the same R-CLI engine, with a clean visual interface, like a blend of VS Code and an AI agent. Either one completes this challenge.

Docs for this whole challenge:

---

Step 0 (Windows only): Set up WSL

R-CLI installs on macOS and Linux. On Windows, you run it inside WSL (Windows Subsystem for Linux), which gives you a real Linux terminal inside Windows. It is free, official Microsoft stuff, and takes about 10 minutes.

1. Click the Start menu, type PowerShell, right-click it, and choose Run as administrator.

2. Run:

   wsl --install
   

3. Restart your computer when asked.

4. After restarting, a terminal window opens and installs Ubuntu. Pick a username and password when prompted (the password will not show as you type, that is normal).

You now have a Linux terminal. Open it anytime by typing Ubuntu in the Start menu. Do the rest of this challenge inside that Ubuntu terminal.

Mac users: open the Terminal app (press Cmd+Space, type Terminal, hit Enter).
Linux users: you know where your terminal is.

---

Step 1: Install R-CLI

Copy this into your terminal and press Enter:

curl -fsSL https://app.backboard.io/api/cli | bash

What this does: downloads Backboard’s official install script and runs it. It takes a minute or two.

If the terminal says curl: command not found (some fresh Linux/WSL setups), install curl first:

sudo apt update && sudo apt install -y curl

then run the install command again.

---

Step 2: Verify it installed

backboard --version

You should see a version number. Also try:

backboard --help

which lists everything the CLI can do. If backboard is “not found,” close your terminal, open a new one, and try again (the install updates your path, and a fresh terminal picks it up).

---

Step 3: Log in

backboard login

R-CLI shows you a URL and a device code. Open the URL in your browser, enter the code, and approve. Your credentials are saved locally, so you only do this once.

---

Step 4: Your first session (the walkthrough)

Start the agent:

backboard

You are now in an interactive session. Try these slash commands, in order:

• /help — shows every available command.
• /model — lets you pick which AI model powers your agent, plus a thinking mode. Remember the model library from Challenge 2? Same idea. Pick something solid for coding.
• /skills — opens the skill picker: pre-built playbooks your agent can follow.
• /mcp — manages MCP servers, which are connections to outside tools and services your agent can use.

Now give it its first real job. Type this (or your own version):

Create a file called hello.py that prints a random programming joke, then run it and show me the output.

Watch what happens: R-CLI writes the file, runs it, and shows you the joke. If the code errors, it reads the error and fixes it itself. That loop—write, run, fix—is what makes it an agent and not just autocomplete.

One more to feel the power:

Make a tiny single-file website called index.html with a countdown timer to the end of Global Hack Week, with fun styling. Then tell me how to open it.

Open the file it made in your browser. You just shipped a website by describing it.

Useful startup flags for later (run these instead of plain backboard):

backboard --model openai/gpt-5.5     # start with a specific model
backboard --thinking high            # more careful reasoning
backboard --memory on                # agent remembers across sessions
backboard --print "your prompt"      # one-off, non-interactive run

---

Alternative: Backboard Studio (desktop app)

If terminals are not your thing, Backboard Studio is the desktop version: the same recursive engine as R-CLI wrapped in a full visual coding environment. Files, editor, and agent chat in one window. If you have used VS Code, Zed, or Cursor, it will feel familiar, except the agent does the building.

1. Download Studio.
2. Install and open it, then sign in with your Backboard account.
3. Do the same walkthrough as Step 4 above: ask it for hello.py, then the countdown website. Same prompts, same results, friendlier interface.

Studio and R-CLI both count for this challenge and both work for Challenges 4 to 6. Pick whichever feels better.

---

How to complete this challenge

Submit one screenshot showing either:

1. Your terminal with the output of backboard --version AND your first session doing the hello.py task, or
2. Backboard Studio open with the hello.py task completed.

[SUBMISSION LINK]

---

Troubleshooting

Problem	Fix
curl: command not found	Run sudo apt update && sudo apt install -y curl, then retry the installer.
backboard: command not found after install	Open a brand new terminal window and try again.
Login URL will not open in WSL	Copy the URL into any Windows browser manually and enter the device code there.
Agent gives an auth or credit error	Run backboard login again, and check your credit balance at app.backboard.io. No credits? Make sure you registered via app.backboard.io/hackathon with code GLOBALMLH2 (Challenge 1).
Windows without WSL	WSL is the supported path. Follow Step 0. If your machine cannot run WSL, use Backboard Studio instead.

---

Next up → Challenge 4: Build a Basic Agent. You have the account, the credits, and the tool. Time to build your first real agent on the Backboard API.