AI
It forgets
Every session starts from zero
Your assistant gets a persistent, structured memory it can navigate from a single entry point, so context is loaded when it matters and never re-explained.
Give your AI a memory.
Give your team a method.
AI-KIDO turns the way you work with AI assistants into a structured, shared and lasting practice: a Memory Bank your LLM can navigate, and a path your team can learn.
- Tree-structured knowledge
- One source of truth
- Works with Claude Code, Zoo Code, Codex, Antigravity and more
- Open source · MIT
AI is powerful. Without a method, it is unpredictable.
Language models can transform how you build. But used on the fly, they forget, contradict themselves and leave each developer with their own habits. AI-KIDO replaces improvisation with a shared, durable way of working.
It forgets
It contradicts itself
Answers that drift over time
Truth tables give every fact one home. When sources disagree, the authoritative one is known and the gap is caught early.
It is a solo sport
Everyone prompts differently
A common language, defined roles and shared workflows let a whole team collaborate with AI the same way.
It does not last
Documentation ages the day it is written
Internal agents validate, rebalance and index your knowledge, so the Memory Bank keeps pace with the project.
How it works
Three ideas, applied consistently. Together they turn scattered notes into a knowledge base that both humans and AI can trust.
-
Structure your knowledge
One tree, one entry point. Knowledge is organized by concepts, techniques, personas and use cases, so an assistant finds exactly what it needs and nothing more.
-
Decide what stays open and what is settled
Truth tables stay flexible and evolve with the project. Hard points lock the decisions you have validated and always win when there is a conflict.
-
Let agents keep it healthy
Internal agents search, check compliance, resolve conflicts, fill gaps and update the base, while preserving your project's own context.
What you can do with it
A complete toolbox around the method, ready to use from your AI assistant.
Memory
A persistent context for your LLM
A hierarchical Memory Bank that compensates for the model's limits and keeps your project's knowledge available across sessions.
Search
Find anything, in docs or in code
Explore the documentation, the source code, or both in parallel, with search plans adapted to your technology stack.
Trust
Know which source wins
Detect conflicts between specifications and reality, and arbitrate with a clear hierarchy of truth.
Quality
Validate as you go
Automatic compliance checks on structure, headers and links, plus a full audit of your base whenever you want one.
Analysis
Understand a project quickly
Analysis plans from a quick overview to an in-depth expert review, and a project index that is cached for speed.
Updates
Stay current without losing your work
Update from the official source and choose per file to replace, merge or ignore. Backups and your local context are preserved.
Team
Organize who does what
Specialized personas and a RACI matrix classify every document and clarify responsibilities across the team.
Integration
Plugs into your AI tools
Built on the AGENTS.md and Skills standards, AI-KIDO works today with Claude Code, Zoo Code, Codex, Antigravity and any other tool that follows them.
Use it where you work
From a weekend project to a multi-project organization, in standard or lightweight mode.
Start a new project
Initialize a Memory Bank in minutes and grow it with the project, from the very first prompt.
Adopt it on an existing codebase
Analyze what is already there, structure it into truth tables and fill the gaps progressively.
Bring your whole team
One shared base, common standards and clear roles, so results stay consistent from one person to the next.
Go beyond code
The method has already been applied to teaching, fiction writing and observability, as well as software.
Philosophy
The name says it all: "AI" and Aikido, the way of the harmony of energies. Instead of fighting the limits of a language model, you learn to use its energy and turn apparent weaknesses into strengths. A small set of fundamental rules keeps the knowledge base robust, scalable and consistent.
🌳 Tree structure
Recursive navigation from a single entry point.
📊 Theory / practice
Clear separation between concepts and techniques, personas and use cases.
🔄 Self-feeding
Continuous improvement based on lessons learned.
⚖️ Balancing
File size constraints and a maximum depth of 4 levels.
🔗 Bidirectional links
Consistent navigation in every direction between documents.
👥 Role distribution
Collaborative organization backed by a RACI matrix.
The real challenge is not learning to use AI, but learning to dance with it. The spirit of AI-KIDO
A method you learn, not just a tool you install
AI-KIDO was born as a way of learning to work with AI, and the repository is the toolbox that grew out of it. You progress through belts, from your first structured prompt to agents that run your production chain and a system that keeps improving itself, with hands-on katas and a dojo spirit of patience and curiosity.
- White belt · Foundations
- Yellow belt · Expansion
- Orange belt · Mastery
- Green belt · Professional practice
- Blue belt · Agentic industrialization
- Black belt · Continuous improvement
Structure
The single entry point is prompts/index.md. Hard points
(hard_points) are authoritative, followed by the Truth Tables (gen).
prompts/
├── index.md # single entry point
├── hard_points/ # immutable truths (absolute priority)
│ ├── agents/ commands/ system/ ...
└── gen/ # project Truth Tables
├── concepts/ techniques/
├── agents/ personas/
└── use_cases/
A Memory Bank that maintains itself
agent-aikido-* agents are tools, agents and modes at the same time.
They automate the maintenance of the Memory Bank so your team can focus on building.
Explore Bank / Code / All
Search the documentation, the code, or both in parallel.
Truth Evaluator
Determines the authoritative source when sources conflict.
Validator & Doc Header
Checks file compliance and manages document headers.
Balancer & Index
Rebalances the tree and keeps indexes up to date.
Bank Populate & RACI Delegator
Fills the gaps and classifies new documents.
Fusion & Update Controller
Updates AI-KIDO while preserving the project's local context.
Get started
Install AI-KIDO in your project or update an existing base. It works for new and existing projects alike.
# Initialize AI-KIDO in the project
/init-aikido
# Update an existing base
/maj-aikido --update
# Compliance audit
/analyse-aikidoSource repository: github.com/lcalmejane/aikido
Ready to work with AI, not against it?
Start with a single prompt and a single entry point. The rest of the method grows with you.