01
Scope before producing.
Define the need, constraints and success criteria. Design the solution around a concrete task.
AI agents.Projectsmoving forward.
An idea to build. A process to improve. Shogun Agents designs AI agents and agent architectures with explicit roles, verification steps and documentation to help you use them.
For individual projects and business needs.
ILLUSTRATIVE DIAGRAM
From intent to deliverable.
01Need & criteria
02Orchestration
03Production
04Review
05Verification
CorrectionDelivery
Explicit rolesTask-specific verificationUseful documentationCustom design
01 / THE APPROACH
A method before
a multitude of agents.
The aim: structure the work so that every role, decision and result can be understood. The number of agents follows from the need.
02
Separate responsibilities.
Give production, review and verification explicit roles. Check review conclusions against verifiable evidence.
03
Make the solution readable.
Explain instructions, mechanisms and design choices. Document how to use the solution and how it can evolve.
Intended design principles. Verification and iteration limits are defined for the chosen task and environment.
02 / THE ARCHITECTURE
Every step has a role.
Every output has a criterion.
Explore an illustrative workflow. Select a step to examine its responsibility and what it passes to the next one.
Correction / Production
RESPONSIBILITYSTEP 01 / 06
Define what success means.
Work starts with an explicit need. Constraints and the evidence required for acceptance are defined before production.
- Input
- Need, context and constraints.
- Decision
- Scope, acceptance criteria and limits on time, cost or iterations.
- Output
- A scoped task and verifiable criteria.
RESPONSIBILITYSTEP 02 / 06
Organize responsibilities.
The orchestrator distributes the work and organizes handoffs between roles. Models and tools are chosen to fit the task.
- Input
- Scoped task and chosen environment.
- Decision
- Required roles, sequence, permitted tools and review points.
- Output
- An explicit work plan.
RESPONSIBILITYSTEP 03 / 06
Produce a first version.
The production role carries out the task within the defined scope. Its result remains a candidate for verification, with the context needed for review.
- Input
- Plan, instructions and permitted context.
- Decision
- Implementation choices within the scope.
- Output
- A candidate deliverable and its explanations.
RESPONSIBILITYSTEP 04 / 06
Compare the result with requirements.
Review examines choices, inconsistencies and gaps against the need. The name ‘reviewer’ alone does not demonstrate the independence or quality of the check.
- Input
- Candidate deliverable and acceptance criteria.
- Decision
- Gaps to address and points to verify.
- Output
- Specific observations for verification or correction.
RESPONSIBILITYSTEP 05 / 06
Decide using verifiable evidence.
Depending on the task: executed tests, file inspection or examination of the result. If a gap is found, production can resume within the limits set during scoping.
- Input
- Result, review observations and criteria.
- Decision
- Accept, correct or stop to reassess the scope.
- Output
- A conclusion with available evidence and limitations.
RESPONSIBILITYSTEP 06 / 06
Provide what is needed to use and adapt.
The deliverable comes with instructions for the chosen environment. Limitations and unverified points are stated explicitly.
- Input
- Accepted result and verification evidence.
- Decision
- What is ready to use and what still needs validation.
- Output
- A deliverable, its documentation and known limitations.
Educational navigation, without agent execution. This diagram does not reproduce a verified Shogun architecture. Measured results will be published after measurement.
03 / THE DELIVERABLES
Files you
can understand.
Readable instructions, an explained architecture and a guide to the chosen environment. Documentation is part of the design.
Behavior
An AGENTS.md defines the agent’s role, instructions and limits.
Operation
From Pro: responsibilities, mechanisms, model choices and explanations of the steps.
Getting started
A README explains usage and, for Pro and Max, how to modify or extend the solution.
EXAMPLE DOCUMENT STRUCTURES
AGENTS.md
# Production agent
## Mission
Carry out the task within the defined scope.
## Responsibilities
- Read the acceptance criteria.
- Explain implementation choices.
- Hand over a result for verification.
## Limits
- Request a decision when scope changes.
- Distinguish produced from verified results.
- Respect the agreed iteration limits.README.md
# Getting started
## Chosen environment
Describe the harness, tools and prerequisites.
## Usage
1. Prepare the required accounts and access.
2. Provide the task and its success criteria.
3. Examine the result and verification.
## Extension — Pro / Max
Explain roles, model choices and the elements
that can be changed or extended.
## Known limitations
State what still needs validation.Educational specimens. These excerpts are not Shogun files, client deliverables or executable configurations.
04 / THE HARNESSES
Your tools.
A tailored design.
-
Claude Code
-
Codex
-
Hermes Agent
-
Kilo
-
OpenCode
-
Gemini CLI
01 / 06Claude Code
Adaptation to be validated
05 / THE OFFERINGS
The right scope
for your needs.
Offerings in development
Two ranges. Three design levels. Custom orders, without a subscription.
Personal — For building your own projects with agents whose roles and operation are documented.
Business — For solutions tailored to business processes, development and operational efficiency goals.
The range adapts the purpose. Levels and deliverables stay the same; pricing will be defined separately.
ONE AGENT’S BEHAVIOR
Plus
One AGENTS.md
Define an agent’s role and instructions within a focused design scope.
- AGENTS.md
- README for the chosen environment
A more limited design scope than Pro and Max. Does not include a complete architecture.
A CUSTOM ARCHITECTURE
Pro
A complete solution
Structure roles, mechanisms and design choices around a specific need.
- AGENTS.md and README
- Explained architecture and mechanisms
- Agent roles and types
- Model recommendations per task
- Rationale for each step
- Guide to modification and extension
A more in-depth design, with the explanations needed to use and extend the solution.
MULTIPLE NEEDS, ONE BUNDLE
Max
Five Pro-level solutions
Organize five custom agents or solutions around your needs, with enhanced follow-up.
- Pro deliverables adapted to each solution
- Separate, coordinated or mixed solutions
- Enhanced follow-up, terms to be defined
The bundle aims to offer an economic advantage. It does not necessarily mean a team of five sub-agents.
ACCOUNTS & USAGE
You provide the accounts and cover model and API usage costs. No credits, AI subscription or model access are included.
EXECUTION ENVIRONMENT
The README fits the chosen harness: the environment that runs and manages agents. Compatibility and prerequisites are to be defined for the need.
06 / INTERNAL WORK
Building for
our own needs, too.
Shogun Agents also plans to use its agents to design and release technology products, including games. This internal work will put the architectures to use on concrete development needs.
A planned activity. No product or result is presented as already available.
07 / USEFUL QUESTIONS
The details matter.
What is defined, what depends on the need and what remains to be clarified.
What does AGENTS.md mean?
It is an instruction document describing an agent’s expected behavior, role and limits. It must be used within an environment capable of running and managing that agent.
Which models and environments are supported?
Models should fit the need and be checked for each order. Supported environments remain to be defined. The site does not announce a fixed model list or universal compatibility.
Are model costs included?
No. You provide your accounts and pay for model and API usage. Services cover design and deliverables; no model access or usage credits are included.
Does Max mean five sub-agents?
Not necessarily. Max bundles five Pro-level agents or solutions: separate solutions, coordinated agents or a mix, depending on the need.
What are the support terms, timelines and prices?
Max includes enhanced follow-up. Its duration, channel and scope remain to be clarified, as do prices, timelines, revisions and maintenance responsibilities. Unlimited support is not promised.
Is there a demonstration or benchmark?
The site presents an approach and educational examples. It does not yet publish a verified demonstration or measured result. Future measurements should state the tasks, comparison baseline, models, budgets and limitations.
Can I place an order today?
Offerings are in development. This site presents the project and the proposed levels. Ordering terms and a verified contact will be added once defined.
ABOUT
Shogun Agents.
A project in development dedicated to AI agents and agent architectures. For individuals building their projects and businesses seeking to structure development and improve their processes.
The ambition: useful agents,
understandable choices.