The CognoKratos Book
Build agents. Share the knowledge.
Engineering Agentic Systems: a living engineering curriculum for building increasingly autonomous systems that humans can trust.
CognoKratos starts from a simple premise: AI agents are moving from systems that answer questions toward systems that act. They will increasingly invoke services, manage resources, prepare financial intent and participate in economic workflows.
That creates a deeper engineering problem than model quality:
How do we give autonomous systems meaningful power without surrendering control over that power?
This book studies that question through five runnable systems and one synthesis part. The model is treated as a probabilistic component inside a larger trust architecture. Identity, capabilities, deterministic policy, durable state, human authority, cryptographic boundaries and audit remain explicit parts of the software around it.
The long-term direction is toward trustworthy autonomous economic agents. Blockchain and cryptography appear where they materially change the trust model: signatures, custody, ownership, verifiable authorization, settlement, programmable constraints and coordination across organizational boundaries.
The current five projects are the present state of the curriculum, not its permanent definition. CognoKratos is designed to grow as engineers contribute new experiments, failure cases, reference architectures and learning tracks.
Read A living curriculum for how the book, repositories and community fit together.
Where to start
- New to AI agents themselves? CognoKratos starts where most agent tutorials stop. Audience and prerequisites suggests where to learn the fundamentals first.
- New to the series? Read What this book teaches, then A living curriculum, and follow the recommended full reading order.
- Have a specific question? Pick an independent route into the part that answers it.
- Want the vocabulary first? Read Models, tools, capability, identity and authority.
- Already using a framework, cloud platform or protocol? Read Where CognoKratos fits to see which of its guarantees the book studies, and which remain yours.
- Comparing designs? Go straight to Part VI.
- Want to improve the curriculum? Read Corrections and contributions.
What this edition covers
The current arc is:
production agent foundations → durable runtime → governed decisions → cryptographic capability → economic intent → trust-architecture synthesis
Each part is built from learning material that lives beside the code in its CognoKratos repository: learning paths, concept chapters, labs, walkthroughs, case studies and challenges. The book adds orientation, transitions between parts and cross-project synthesis. Every imported chapter names its repository and pinned revision.
Important frontiers remain open by design, including secure signing, settlement, on-chain execution, portable agent identity, delegated authority, smart-contract controls and multi-agent economic coordination. These are research directions for future contributors, not claims that the current edition already solves them.
What this book is not
It is not a framework, a product manual, a compliance guide or a claim that one architecture is universally correct. Nor is it an introductory agent course, a governance product, an identity protocol, a wallet SDK or a payments network: those exist, and Where CognoKratos fits points to good ones. The book's subject is the architecture between them.
The systems are reference architectures for learning. They are explicit about what they implement and what they leave out. At the revisions used by this edition, none of them executes trades, moves money or signs transactions. Each part's introduction states what its project actually does, and Limitations and maturity collects the gaps in one place.
Reference architectures are propositions to inspect, test, break and improve.
CognoKratos is an open-source, community-built engineering curriculum supported by BelaZayka GmbH. cognokratos.com