Skip to content

Getting set up in VS Code

This is the one page that assumes nothing. If you have never opened a code editor in your life, follow it top to bottom and you'll end up in the same place every engineer on this project works: the repo open on your laptop, with Claude Code sitting beside it, ready to answer a question or make a change. You won't write a line of Python — you'll talk to the code.

By the end of this page you'll be able to

  • Open the klaus-u2v/automations repo on your own computer.
  • Ask Claude Code, in plain English, to explain any part of the system.
  • Ask it to change something, or to build a whole new automation — and understand the two settings that make those answers good.

It's a one-time setup of roughly 15–20 minutes. You only do it once per computer.

What VS Code is — and why we use it

VS Code (Visual Studio Code) is a free editor from Microsoft. It's the industry-standard place to open a codebase — think of it as the "Microsoft Word for code," except we barely use it to type. For us its real job is to be the home of Claude Code: the agentic assistant that actually reads and writes this repo on your behalf.

That's the mental shift worth making up front. You are not here to learn Python. You're here to drive an agent that already knows Python — and, thanks to the CLAUDE.md files in every project, already knows this system. VS Code is simply the window you do that through.

Three things meet inside that window, each from an earlier step below:

flowchart LR
    You["👤 You<br/>plain-language requests"] --> VSC
    VSC["💻 VS Code + Claude Code<br/>on your laptop"] -->|reads &amp; edits| Repo["📁 Your copy<br/>of the repo"]
    VSC -->|reasoning| Claude["🤖 Your U2V<br/>Claude account"]
    Repo -.->|clone / push| GH["🐙 GitHub<br/>klaus-u2v/automations"]

Workbench, not runtime

VS Code is where the system is built and changed — it is not what runs it. The live automations run on GitHub Actions on a schedule (see Structure & orchestration). Nothing you do on your laptop affects the live system until your change is merged into main. That makes this a safe place to explore.

Before you begin

You need two accounts. Both are things someone at U2V grants you — set them up before you start so the steps below don't stall:

Account What it's for How you'll use it
A GitHub account Holds the code; lets you download it and save changes back Sign in once inside VS Code (Step 2)
Your U2V Claude account Powers Claude Code's reasoning Sign in once inside VS Code (Step 3)

Needs your detail — who provisions access

Add here: who at U2V invites a new person to the klaus-u2v GitHub organisation (so they can see the repo), and who adds them to the U2V Claude for Work team (and on which plan). These are the two "ask an admin" prerequisites; everything else on this page is self-serve.

Step 1 — Install VS Code

  1. Go to code.visualstudio.com and click the big Download button. The site detects your operating system automatically (macOS or Windows).
  2. Install it like any other app:
    • macOS — open the downloaded .zip, then drag Visual Studio Code into your Applications folder.
    • Windows — run the downloaded installer and accept the defaults (leave "Add to PATH" ticked).
  3. Open VS Code. On first launch it offers a colour theme and a few options — pick a theme and skip the rest; none of it matters here.

Step 2 — Sign in to GitHub

GitHub is where the code lives. Signing in lets VS Code download the repo and, later, save your changes back to it.

  1. In VS Code, click the Accounts icon — the small person/silhouette icon at the bottom of the narrow strip on the far left (the Activity Bar).
  2. Choose Sign in with GitHub. Your browser opens.
  3. Log in to GitHub and click Authorize to connect it to VS Code, then return to VS Code.

If you skip this, you'll just be asked later

VS Code will also prompt you to sign in to GitHub the moment you try to clone in Step 4 — so if nothing happens here, don't worry, it comes back around.

Step 3 — Install the Claude Code extension

This is the piece that turns VS Code into your workbench. Keep it simple:

  1. Open the Extensions view: click the icon that looks like four squares in the Activity Bar, or press Cmd+Shift+X (macOS) / Ctrl+Shift+X (Windows).
  2. In the search box, type Claude Code. Pick the result published by Anthropic.
  3. Click Install.
  4. Once it's installed, open the Claude Code panel by clicking the Claude icon in the Activity Bar (the far-left strip). A sign-in screen appears.
  5. Click Sign in → your browser opens → sign in with your U2V Claude account (the one U2V added you to) → return to VS Code.

That's it — no API key, no configuration. You're connected.

If the panel doesn't appear

Open the Command Palette (Cmd/Ctrl+Shift+P), type Developer: Reload Window, and run it. That resolves almost every "I installed it but don't see it" hiccup.

Step 4 — Clone the repo

"Cloning" just means downloading your own working copy of the code to your laptop.

  1. Open the Command Palette: Cmd+Shift+P (macOS) / Ctrl+Shift+P (Windows).
  2. Type Git: Clone and press Enter.
  3. Paste the repository address: 
    https://github.com/klaus-u2v/automations.git
  4. Choose a folder on your computer to keep it in (e.g. a Projects folder). VS Code downloads the repo.
  5. When it finishes, VS Code asks whether to open the cloned repository — click Open.

You're now looking at the whole codebase: the u2v-* project folders, the docs/ you're reading now, the CLAUDE.md files. You don't need to understand any of it yet — Claude Code will.

Step 5 — Open your first Claude chat

  1. Click the Claude icon in the Activity Bar to open the panel (or use the Command Palette and type Claude Code).
  2. Start a new chat.

The chat box is now pointed at the entire repo you just opened. Because every project ships a CLAUDE.md operating guide, Claude Code already has the context to answer well — it isn't reading this system cold.

A good first question

Before anything else, ask it to orient you:

"Give me a tour of this repo — what are the projects, and how do they fit together?"

It'll read the CLAUDE.md files and walk you through the whole system. That single answer is the fastest way to make the rest of this site click.

Step 6 — Ask, propose, build

From here you work entirely in plain language. There are no commands to memorise. The three things you'll do most:

  • Understand — "Explain how u2v-dealflow scores a startup." · "How does lp-share decide which deals count as new?"
  • Change — "Add a new sector to the dealflow scoring rubric." · "Change how often the dispatcher checks the inbox."
  • Build — "Scaffold a new automation that does X." Adding one is deliberately cheap; the shape is identical every time (see Adding an automation).

The two dials that change your answers

You can get real work done knowing just two controls. Both sit in the row along the bottom of the chat box, and both are worth understanding because they visibly change the quality of what you get back.

1. Mode — brainstorm first, or make changes now

By default Claude will go ahead and make edits (Edit automatically). When you'd rather think before touching anything — scoping a new project, weighing an approach, or just understanding the codebase — switch the mode to Plan.

Click the mode indicator at the bottom of the chat box to switch it.

Why Plan mode gives better answers

In Plan mode, Claude first explores your codebase — reading the relevant files — and only then answers or lays out a step-by-step plan. It changes nothing until you approve. That exploration is exactly why its answers about the current technical stack are more informed: it's reasoning from what the code actually says, not from a guess. When the plan looks right, you approve it and let it proceed.

Mode What it does Reach for it when
Plan Explores & proposes; makes no changes until you approve Brainstorming, scoping, understanding — anything exploratory
Edit automatically Applies edits as it works You already know what you want done
Manual Asks before each individual edit or command You want to approve every step as it goes

2. Effort — how hard it thinks

Claude Code lets you dial its reasoning effort up or down. The levels run Low → Medium → High → Extra High → Max. More effort means it thinks longer and harder before answering — slower, but noticeably better on complex work.

Rule of thumb

For a quick question, the default is fine. For a bigger project — designing a new automation, a tricky change across several files — set the effort to Extra High or Max. Pair it with Plan mode and you get the system's best thinking: deep exploration, then a considered plan, before a single change is made.

Where this fits

You now have the workbench. To go deeper:

  • Design principles — the seven ideas that shape every project; read this and the codebase stops looking like seven separate things.
  • Structure & orchestration — the external tools that run the live system (GitHub Actions, the Claude platform, Gmail, the cron trigger) and where their settings live.
  • The automations — a deep-dive on each of the seven projects.

And remember the loop: explore and change all you like on your laptop — it only becomes live once the change is merged into main.