Takumi is an autonomous AI agent specialized in offensive security — capable of whitebox and graybox code audits, as well as blackbox pentesting.

We began this project with a simple belief:

Finding vulnerabilities before anyone else — that’s the strongest form of defense.

Before Takumi, we had already spent years conducting deep human-driven research on software security — contributing to the safety of major products such as Unity, git, Canonical LXD, OpenWRT, Ubuntu, language ecosystems (known as "BatBadBut"), etc.

Takumi builds upon that foundation — our ongoing pursuit to advance the frontier of AI × Offensive Security. Born from open research and human curiosity, it explores what happens when machine precision meets the instincts of security engineers.

We’re GMO Flatt Security — a Tokyo-based team of security engineers, researchers, and builders working to protect Japan and the world. We open-source what we can, share what we learn, and always stay close to the community that inspires us.

• 💬 Join our free beta program: contact [email protected] (our CTO will respond to your email)
• 💬 Buy the stable version now: cloud.shisho.dev/hello/takumi

• 🔬 Our latest research: flatt.tech/research
• 🧠 The story behind Takumi: flatt.tech/en/takumi
• 🌐 flatt.tech/en
• 🐦 @flatt_security_en (in English) / @flatt_security (in Japanese)