Claude Code

This is an Agentic Coding Tool that reasons and writes code.

To be, most important is still the improved AI workflow, not the autocomplete of AI-based IDE, and that’s where Claude Code is doing everything right. With it’s powerful Agentic Coding Tool style, it’s super powerful to create todo list for itself, interpreting existing code and working its way throught it, autonomously.

As I use the cmd line extensively, with Tmux and Neovim, it’s just a joy to use Claude Code, as it’s implemented with such details within the cmd line. Perfect UI with possibly something like Charm Shell included.

As said in GenBI, it’s important to utilize speed, integrate MCPs and also solving the right problems. To me, Anthropic with Claude Code seems to do many things right.

Also you can follow each step its doing, and get full control on each prompt, e.g. not accepting a change, or telling it even why that was wrong.

Claude code takes advantage of the Unix Philosophy, and uses these tools autonomous to run and debug errors. This is how I use claude code all day, to add a small feature to my website, or analyze my errors in my linux setup on Omarchy.

Also, everything of my value is locally, my Obsidian vault (although I don’t use claude code for that, as I like my stuff to be private), my websites, my book, everything is in simple Plain Text Files, so claude code can help with all these things. This is even more true with MCPs that connect to your database (MotherDuck MCP), BI tools (Using Rill MCP), or any other integration.

Here’s a longer write-up, more focused on Obsidian, great read: The Magic of Claude Code.

And the best of all, AI and its AI Agents works best on Plain Text File. So if you store your notes in something like Obsidian, all of a sudden you can inject all your knowledge on demand to the AI whenever needed.

As I do store all my information such as journalling, new ideas, work, everything in my Second Brain, this gets a really powerful tool whenever I need. I never point all notes to Claude tought, as I have sensitive data, and I still want to be in the driver seat, but I reguarly share a couple of notes that fit the context, and it will do a much better job than doing my prompt with no further context.

My first try and impression of Claude code I share on YouTube with dlt and Snowflake: First Impressions as a Data Engineer - YouTube.

I used Claude Code for my Markdown grammar checks and fixes.

I created a project fix-grammar.md that has a prompt how to fix my text (without changing words or major changes):

The changes look like this:

As everything is in git, I can simply see the diffs after Claude is finished.

The summary what he did looks like this:

I can re-run that and even have different prompts.

This was the cost for running it once:

Read more on AI Writing or discuss on Bsky.

• Vim Motions can be activated
• / for special operation. E.g. /init for creating a claude.md. There many more.
• CLI reference - Claude Code Docs
• Interactive mode

You can switch into Plan Mode during a session using Shift+Tab to cycle through permission modes.

Plan Mode instructs Claude to create a plan by analyzing the codebase with read-only operations, perfect for exploring codebases, planning complex changes, or reviewing code safely.

Plan mode will make it create a plan for itsself, and along the way using multi-agents. Meaning there’s multiple tasks running now not only one that called claude:

Same as in my terminal where I use Recursive Search in Terminal with fzf, claude has this built in:

SDK - Anthropic

1
2
3
4
5
6
7
8

# Basic syntax

claude mcp add <name> <command> [args...]

# Example: Adding a local server
claude mcp add my-server -e API_KEY=123 -- /path/to/server arg1 arg2
## Claude Desktop
Claude desktop is the windows or MacOS version. 

see Model Context Protocol (MCP) - Anthropic. My configs in Claude Configs (MCP).

See Claude Configs (MCP).

Announed 2025-10-20, Claude Code Web, see Claude Code on the web.

Try it at claude.ai/code/.

• How Claude Code Creator uses Claude - Claude Code creator Boris shares his setup with 13 detailed steps,full details below

Origin: Announced 2025-02-25: Claude 3.7 Sonnet and Claude Code \ Anthropic
References: Anthropic, Claude Configs (MCP)
Created 2025-02-26