KaKeKiKoKu Studio (Knowledge Engine) is a tool that captures and preserves the decisions, constraints, rejected alternatives, and human intention that get lost when you work with AI. It lives alongside your project files and gives your AI the context it needs to work as if it had been there from the start.
I am a developer with over 40 years in the software industry. I have worked in every role from junior programmer to software architect, always in enterprise software — corporate products, intranets, complex systems for large organizations.
After decades building software for others, I felt the need to create something of my own. Something that fellow professionals would use. Something creative, but grounded in the deep knowledge of how software is really built.
KaKeKiKoKu was not born from a brilliant idea. It was born from failing.
I started with vague, impossibly diffuse ideas about what working with AI should feel like. I experimented with dozens of approaches. Each one seemed to work on paper but collapsed in practice. Ten complete redesigns. Five times I threw away the code and started over. There were moments when I thought I was wasting my time — that what I was building was absurd, that a plain text file was good enough.
But I kept going. And through repeated failure, iteration after iteration, KaKeKiKoKu started to work. Not because I got it right, but because I got it wrong enough times to finally understand the problem.
KaKeKiKoKu was first built in close collaboration with Claude Opus, which acted as an active development partner throughout its creation. Claude was not just a coding tool — it was part of the design process. It gave opinions, built prototypes, tested the product from the AI side, and reported friction. Many of KaKeKiKoKu's design decisions emerged from conversations between a human and an AI working together on the same problem.
That origin matters, but it is not a lock-in strategy. KaKeKiKoKu opened the same Knowledge Engine core to MCP-capable workflows beyond where it started: Codex, Antigravity, Claude Desktop Code tab, and generic MCP clients. Claude Code still has the deepest automatic integration today; the core idea is broader than any single agent.
The result is a product that understands both sides: what the human needs to manage and what the AI needs to receive to do its best work. KaKeKiKoKu is for both.
KaKeKiKoKu is an independent product with no commercial, organizational, or development affiliation to Anthropic or any AI provider. The relationship described above is that of a user and a tool.
KaKeKiKoKu has been in continuous development for a year, shipping release after release since April 2026. The early versions were honest alphas; the Beta label arrived only after KaKeKiKoKu was tested outside its own codebase — fresh AIs using it blind, real projects that were not KaKeKiKoKu itself — and the bugs that round surfaced were hunted down one by one. Today it runs on Windows and Linux, speaks six languages, and teaches itself through a built-in interactive tutorial. macOS is planned.
AI has made it easy to ship an app in a weekend. KaKeKiKoKu is not that. It is a year of thousands of prompts, tests, dead ends, and redesigns — directed by a human who knew what he was looking for. The current release is a deliberate stabilization point: feature work is paused in favor of polish, while a long list of ideas waits for its turn.
The core of KaKeKiKoKu — knowledge management and context — is free, and I have no plans to change this. After 40 years in this industry, this is my way of giving something back. KaKeKiKoKu is built for professionals who work intensely with AI, and I believe the tool that makes their AI smarter should not be behind a paywall.