# Serkan Korkut > Serkan Korkut is a builder and entrepreneur who makes developer tools and products he likes, working with AI native systems and pipelines, agent orchestration, and distributed systems. His backend toolkit includes Java, Spring Boot, Kafka, RabbitMQ, Redis, AWS and OpenTelemetry. This is his personal site. Serkan Korkut builds with an entrepreneurial mindset: find a real problem, turn an idea into something useful, and keep learning from the people who use it. He builds developer tools that remove friction for engineers and products he would use himself. His other areas of focus are AI native systems and pipelines (product workflows with AI at the core), agent orchestration (connecting agents, tools and workflows to cut manual work) and distributed systems (service boundaries, messaging, reliability and design trade-offs). ## Who is Serkan Korkut? Serkan Korkut is a software engineer and builder based in Ankara, Türkiye, with more than ten years of experience in backend and distributed systems. He makes developer tools for AI coding agents, including tokenmeter and mask2ai, and writes about AI agents and engineering on Medium. Türkçe: Serkan Korkut, Ankara'da yaşayan, backend ve dağıtık sistemlerde on yılı aşkın deneyimi olan bir yazılım mühendisi. Yapay zekâ kodlama ajanları için tokenmeter ve mask2ai gibi geliştirici araçları üretiyor. ## Pages - [Home](https://serkan.fyi/): overview, current focus, stack and selected work - [About](https://serkan.fyi/about/): who Serkan Korkut is, how he works and what he focuses on - [Serkan Korkut kimdir?](https://serkan.fyi/tr/): the same profile in Turkish - [Work](https://serkan.fyi/projects/): mask2ai, tokenmeter, and public code - [Writing](https://serkan.fyi/blog/): technical articles - [Back to basics: what Joel Spolsky said in 2001 still decides who can use AI well](https://medium.com/@serkankorkut/back-to-basics-what-joel-spolsky-said-in-2001-still-decides-who-can-use-ai-well-b16f685e3faf): why computer science fundamentals matter more when AI writes the code. November 2026. - [AI: From zero to hero](https://medium.com/@serkankorkut/ai-from-zero-to-hero-ea1877e6df1b): seventy years of AI in ten milestones, the move from chat to agents, tools, MCP, connectors, skills and plugins, and a first automation. November 2026. - [No-AI days: use AI heavily, but keep your brain in the gym](https://medium.com/@serkankorkut/no-ai-days-use-ai-heavily-but-keep-your-brain-in-the-gym-8f033bfaa3eb): why to keep one day a week without AI so core engineering skills do not fade. October 2026. - [Harness engineering in under 3 minutes](https://serkan.fyi/harness-engineering/): article on making AI coding agents reliable - [Experience](https://serkan.fyi/experience/): areas of practice, with a link to LinkedIn for professional background - [Speaking](https://serkan.fyi/speaking/): technical talks and meetup inquiries - [Contact](https://serkan.fyi/contact/): how to reach Serkan Korkut ## Products - [mask2ai](https://mask2ai.com): a Claude Code plugin and Chrome extension that mask personal data (names, emails, phone numbers, cards, IBANs, national IDs) before it leaves your machine for an AI, and restore it on screen. No server, no account. - [tokenmeter](https://tokenmeter.fyi): a local, zero-dependency dashboard that reads the logs Claude Code, Codex and Copilot CLI already keep, showing real cost per prompt, cache misses and rate-limit windows. Install via PyPI (`tokenmeter-dashboard`) or Homebrew (`serkankorkut/tap/tokenmeter`). ## Writing - [Harness engineering in under 3 minutes](https://serkan.fyi/harness-engineering/): what harness engineering is, the agent loop, and six rules that make AI coding agents reliable. Full version: [Harness Engineering Deep Dive](https://medium.com/@serkankorkut/harness-engineering-guide-ai-coding-agents-0b4d60a70bab) on Medium. October 2026. - [What your AI assistant actually receives, and how to keep personal data out of it](https://medium.com/@serkankorkut/what-your-ai-assistant-actually-receives-and-how-to-keep-personal-data-out-of-it-579ffeb05093): how mask2ai keeps personal data out of prompts sent to Claude and ChatGPT. September 2026. - [Tokenmeter: Know what every prompt costs you in Claude Code, Codex and Copilot](https://medium.com/@serkankorkut/tokenmeter-know-what-every-prompt-costs-you-in-claude-code-codex-and-copilot-67c8032a244e): why token counts hide real prompt cost, and how tokenmeter shows it from local logs. September 2026. - [Capturing your screenshots with Foresight in action](https://medium.com/@serkankorkut/capturing-your-screenshots-with-foresight-in-action-8b184a44a368): a practical walkthrough of test screenshots and CI visibility with Thundra Foresight. Originally published on the Thundra blog, December 2021. ## Code - [thundra-foresight-screenshot-capturing](https://github.com/serkankorkut/thundra-foresight-screenshot-capturing): a public fork of Thundra's screenshot-capturing example, accompanying the article above. Presented as a sample, not an independently authored product. ## Elsewhere - LinkedIn: https://www.linkedin.com/in/korkutserkan/ - GitHub: https://github.com/serkankorkut - Medium: https://medium.com/@serkankorkut