Watch the video with chapters & transcript →
AI TOOLS
In short: Anthropic Academy's free "Claude Code in Action" course explains how a coding assistant turns a text-only model into a tool that reads files, plans, and acts.
Anthropic Academy hosts official, free courses on Claude, and "Claude Code in Action" is the one Ani Björkström, a tech consultant based in Stockholm, works through in this tutorial. It starts with a question most users never ask: how does a model that only reads and writes text open your files?
Large language models like Opus, Sonnet, or Haiku can only take text input and return text output; they cannot open a document on their own. The coding assistant gathers context, formulates a plan, and takes action on the model's behalf.
The course's example: if you ask what code is in main.go, the assistant appends an instruction like "if you want to read a file, respond with read file, name of the file." Claude answers with that request, the assistant executes it, and the contents come back as text. Every tool use follows this pattern.
Yes, because Claude Code never indexes your code base or stores chunks of it on a supplier's side. Tools like GitHub workspace typically index all the code in your repository and save those chunks with the vendor, so your entire code base leaves your control.
Claude Code instead runs general commands such as find locally on your computer and sends only the relevant piece to the Claude API. According to the course, this selective retrieval is a key security argument for the approach.
Installation is one command that differs by operating system: macOS, Linux, and WSL share one, while Windows PowerShell and Windows command prompt each use their own, the latter via curl. On first run of the claude command, you pick a terminal color and authenticate with your Claude AI credentials; if the command is not found or you hit network or permission errors, the docs' troubleshooting section covers it.
Start every existing project with the /init command, which reads the whole code base and generates a CLAUDE.md file summarizing the project. From there, the course distinguishes three memory files with different scopes, editable directly or via the /memory command inside Claude Code.
| File | Scope | Typical content |
|---|---|---|
| CLAUDE.md | One project, created by the /init command | A generated summary of the files and structure of that project |
| CLAUDE.local.md | One project, never shared with other engineers | Personal instructions and customizations specific to you |
| Global CLAUDE.md | Every project on your machine | Rules Claude should always follow, such as "use comments sparingly, only comment complex code" |
When you already know where the answer lives, tag the file with the @ symbol followed by the file path, for example when asking how the auth system works. Referencing the exact file spares Claude Code from searching the whole project.
Yes. It is one of the free official courses on Anthropic Academy; you log in, choose the course, and follow along at your own pace.
Yes. Add an instruction like "use comments sparingly, only comment complex code" to CLAUDE.md, in your editor or via the /memory command.
Read the full video transcript, with chapters, on the video page →
Want this working inside your finance team?