Wednesday, November 18, 2026

LLM needs to talk with Spring app

 MCP Server kurmak, bir Yapay Zeka modelinin (örneğin Claude'un) senin verilerine ve dış sistemlerine güvenli ve kurallı bir şekilde bağlanmasını sağlayan standart bir köprü (adaptör) ayağa kaldırmaktır. Nasıl ki veritabanına bağlanmak için JDBC standardı varsa, LLM'lerin dış dünyayla konuşması için de Model Context Protocol (MCP) bir standarttır.

Kafandaki "Kürbnetes MCP server hazır vardır da benim Spring app bana özel, onu nasıl yapacağız?" ikilemini ve işin mekaniğini şu şekilde özetleyebiliriz:

1. Hazır / Standart MCP Server'lar Nedir?

Kubernetes, GitHub, PostgreSQL veya dosya sistemi gibi genel geçer teknolojiler için dünyada birileri (Anthropic veya açık kaynak topluluğu) zaten genel amaçlı MCP sunucuları yazmıştır. Bunları doğrudan alır (örneğin bir npx komutu veya Docker container olarak) ve Claude Desktop gibi bir istemciye yapılandırıp geçersin. Bunlar evrenseldir, herkese açıktır.

2. Senin Spring Boot Uygulaman (Bana Özel)

Senin şirket içi iş kurallarını barındıran, kendi PostgreSQL veritabanına sorgu atan veya özel mikroservislerinle konuşan backend uygulamanı dış dünyadaki hazır bir K8s MCP server bilemez. Claude'un senin özel iş mantığını çalıştırabilmesi için, senin Spring Boot uygulamanın kendisini bir MCP Server'a dönüştürmen gerekir.

3. Kendi Spring App'ini Nasıl MCP Server Yaparsın? (Framework Desteği)

Artık bu iş için tekerleği yeniden keşfetmeye gerek yok. Spring AI kütüphanesi, Java dünyasında tam bu iş için yerleşik MCP Server desteği (spring-ai-starter-mcp-server-webmvc vb.) sunar.

Senin kod tarafında yapman gerekenler sadece şunlardan ibarettir:

  1. Bağımlılığı Ekle: Projene Spring AI MCP Server starter'ını ekle.
  2. Metodu İşaretle (@Tool): Spring Service katmanındaki herhangi bir metoda @Tool annotasyonu koyarsın:
    @Service
    public class OrderService {
    
        @Tool(name = "get_customer_orders", description = "Fetches active orders for a given customer ID from the database")
        public List<Order> getCustomerOrders(String customerId) {
            // Kendi özel business logic'in ve DB sorgun
            return orderRepository.findByCustomerId(customerId);
        }
    }
    
  3. Protokolü Yönetir: Spring AI arka planda JSON-RPC protokolünü, SSE (Server-Sent Events) veya HTTP transport katmanını yönetir. Claude (veya herhangi bir MCP Client) senin uygulamana bağlandığında, hangi araçların (tools) olduğunu otomatik olarak keşfeder (tools/list).

Kullanıcı Claude'a "Serkan'ın son siparişleri neler?" diye sorduğunda, Claude arka planda senin Spring uygulamandaki getCustomer_orders metodunu JSON-RPC ile tetikler, veriyi alır ve kullanıcıya anlaşılır dille raporlar. Yani kendi uygulamanı birkaç satır annotasyonla birer AI aracına (Tool) dönüştürmüş olursun.

Model Context Protocol with Spring AI tutorial Bu video, Spring Boot uygulamalarını MCP sunucusuna dönüştürme sürecini pratik bir örnekle görsel olarak anlaman için faydalı olacaktır.

Sunday, November 1, 2026

SkillSpector

I rarely do this..

Below content is from X

--

 Installing a skill you found on the internet into your agent means running someone else's script on your own machine.

​Nvidia open-sourced SkillSpector for this exact reason. It scans the skill before installation, checking for prompt injection, data exfiltration, and problematic code, and ultimately assigns a risk score.

​With --no-llm, it doesn't even require an API key.

​I think the very first thing to do is check the ones already installed:

​skillspector scan ~/.claude/skills --recursive --no-llm

​If your skill folder is cluttered, run it once today, and make it a habit before installing any new skills so you can have peace of mind.


https://github.com/NVIDIA/SkillSpector

Thursday, March 5, 2026

Daily life Linux problems..

 ## Confessions of a Distracted Linux Coder: How "Port Already in Use" Turned into Heimdall


Hey everyone,


You know those moments when you're deep in code on Linux and suddenly…  

**"Why can't I delete this file?"**  

**"Which damn process is holding port 8080?"**  

**"Wait, who just opened that socket?"**


One minute you're writing a function, the next you're running `lsof`, `ss`, `fuser`, `ps aux | grep`, `journalctl -u`, and `iotop` like a madman.  

By the time you fix the tiny problem, you’ve completely forgotten what you were actually working on.


**"What was I doing again?"** 😅


Or worse — that paranoid feeling creeps in:  

"Is everything okay? Why is systemd-journald eating CPU?  

Who is scraping my disk like it's a frying pan?  

Why are the modem lights blinking non-stop?  

Who is sending all these packets?!"


And the worst one:  

"Did that innocent little package I just installed silently start talking to the outside world without telling me?"


Switching between five different TUIs and terminals just to answer basic questions gets exhausting.  

I got tired of it.


So… I ended up going down the rabbit hole.  

What started as a quick hack to find "who's using this port" slowly grew into something bigger.


And because I can't help myself, I added way too many icons to make it feel a little more fun 😄


Ladies and gentlemen, meet **Heimdall** – my little terminal watchdog that's still very much in its crawling baby stage.


https://github.com/sunels/heimdall


It's nowhere near perfect, but it already saves me a surprising amount of time every day.  

If you ever feel the same frustration, give it a try.  

Feedback, suggestions, or even a tiny PR would genuinely make me happy.


Thanks for reading,  





Sunday, February 22, 2026

Heimdall – How a Simple Port Error Turned into My Crazy Terminal Side Project

 

Hey folks,

Everything started because I got so tired of that classic Spring Boot error: “Port already in use”. Every time I had to hunt down the PID in the terminal, kill it manually, and pray it didn’t happen again five minutes later… ugh.

I was fed up scrolling through ss, netstat, lsof, ps — all that noise just to free a port during a busy workday.

Then things got out of hand (in a good way 😄).

I started building something small… and it kept growing. Now Heimdall is this beautiful curses-based TUI that watches ports, processes, files, risks — and even runs as a daemon in the background.

The coolest parts right now are Sentinel (it spots shady behavior like backdoors, masquerading processes, deleted binaries, suspicious listeners) and Daemon mode (it quietly monitors suspicious outbound connections, suspends the process, and asks you to allow or kill — either via a popup modal if the TUI is open, or a notification + timeout if not).

Whenever I think “hey, this would be cool to add”, I throw it at an AI agent while I still have tokens left 😂. Sometimes I just feel like the product owner who never stops adding features.

If you want to join the fun, check out the repo: https://github.com/sunels/heimdall

or https://pypi.org/project/heimdall-linux/

Give it a star, try it, break it, send feedback/PRs — I’d love to hear what you think (or what crazy feature I should add next).

Thanks for reading, stay safe out there!

Love, Serkan ❤️


Saturday, January 17, 2026

Linux terminal tool for port process open files

 Interactive terminal-based port, process, file, and resource inspector for Linux


portwitr-interactive is a high-performance, curses-based Terminal User Interface (TUI) designed to give you instant visibility and control over your Linux system — all from a single, interactive view.

Tool for sysadmins and developers

✨ Features

🔍 Live port listing using ss

⚡ Shows CPU% / MEM% usage per process

🧠 Maps PORT → PID → PROGRAM

⛔ Firewall toggle for selected port (temporarily block/unblock traffic)

📂 Displays all open files of the selected process (/proc/<pid>/fd)

🧾 Deep inspection via witr --port

🖥️ Fully interactive terminal UI (curses)

⚡ Real-time refresh

🛑 Stop a process or systemd service directly from the UI (with confirmation)

📝 Warnings annotation (e.g., suspicious working directory is flagged but explained)


Checkout and Play: https://github.com/sunels/heimdall

Installation: https://github.com/sunels/heimdall?tab=readme-ov-file#-installation

    wget https://github.com/sunels/heimdall/releases/download/v0.1.0/heimdall_0.1.0-1_all.deb
    sudo dpkg -i heimdall_0.1.0-1_all.deb
    Run:
    sudo heimdall
    
    # or just
    heimdall