Skip to content

Introduction

OpenLogos is an open-source software engineering methodology for the AI era. It turns AI coding tools — Claude Code, OpenCode, Cursor, and others — from guessing machines into precision instruments by providing structured, explicit context at every stage of development.

When you tell AI “build a login feature” without context, every decision is a guess: requirements, UI design, API structure, edge cases, error handling. With 10 decisions per feature, that’s 2^10 = 1,024 possible paths — most of them wrong.

This is Vibe Coding: fast to start, painful to finish. Code gets generated quickly but is riddled with assumptions, missing edge cases, and undocumented decisions. Documentation drifts from reality. Tests are an afterthought. Changes break things nobody remembers.

OpenLogos provides AI with the full picture — requirements, design, API specs, test cases — so it generates code that’s correct, tested, and traceable. Same documents, same context, consistent results.

The methodology rests on four core principles:

1. WHY → WHAT → HOW (Three-Layer Progression)

Section titled “1. WHY → WHAT → HOW (Three-Layer Progression)”

Every project progresses through three phases where each phase’s output feeds the next:

  • Phase 1 · WHY — Requirements: personas, pain points, scenarios (S01, S02 …), acceptance criteria
  • Phase 2 · WHAT — Product Design: feature specs, interaction flows, HTML prototypes
  • Phase 3 · HOW — Implementation: architecture → sequence diagrams → API/DB → test cases → code → verify

Quality gates between phases prevent ambiguity from accumulating. Learn more →

Every feature is decomposed into scenarios. One scenario ID (S01) traces from the first requirement to the last line of verified code — no traceability matrix needed.

Tests are designed before code: unit tests, scenario tests, and orchestration tests. AI generates code against a precise, verifiable target — “pass these 12 test cases,” not “build a login feature.” Learn more →

Every decision lives in a document inside the logos/ directory — Markdown, YAML, JSON. Reviewable, reproducible, cumulative. AI reads the docs, not your mind.

AGENTS.md acts as the AI navigator: it detects the current phase, suggests the next step, and loads the right Skill automatically. Learn more →

Not invented from scratch. Built on 40+ years of proven theory — BDD, TDD, DDD, Stage-Gate, Docs-as-Code — recompiled for AI execution. Learn more →

You + AI coding tool
AGENTS.md ← AI reads this first
Detect phase ← Scans logos/resources/
Load Skill ← e.g. prd-writer, api-designer
Follow process ← Skill provides step-by-step instructions
Produce artifact ← requirements.md, api.yaml, tests, code
  1. You run openlogos init to create the project structure
  2. Open the project in your AI coding tool (Claude Code, OpenCode, Cursor, etc.)
  3. AI reads AGENTS.md, detects the current phase, and loads the appropriate Skill
  4. You collaborate with AI through each phase — it follows the Skill’s structured process
  5. Each phase produces artifacts that become context for the next phase
  6. openlogos verify validates the final result against acceptance criteria

Skills are platform-agnostic Markdown files (SKILL.md) that give AI operational instructions for each task. OpenLogos includes 13 built-in Skills:

PhaseSkills
WHY (Requirements)project-init, prd-writer
WHAT (Design)product-designer
HOW (Implementation)architecture-designer, scenario-architect, api-designer, db-designer, test-writer, test-orchestrator, code-implementor, code-reviewer
Cross-phasechange-writer, merge-executor

Skills work with any AI tool that can read project files. No vendor lock-in. Full Skills Reference →

The openlogos CLI provides 8 commands for project lifecycle management:

CommandPurpose
initInitialize project structure
syncSynchronize AGENTS.md with config
statusShow current phase and progress
verifyValidate test results against acceptance criteria (Gate 3.5)
launchActivate change management after verification passes
changeCreate a change proposal (delta)
mergeMerge approved changes into main artifacts
archiveArchive completed change proposals

Full CLI Reference →

After the initial build, every iteration follows a structured change process:

  1. openlogos change <slug> — create a proposal with impact analysis
  2. Write proposal.md (what changed and why) and tasks.md (phase-based checklist)
  3. Produce delta files for affected artifacts
  4. openlogos merge — apply deltas to main artifacts
  5. Implement code + run tests
  6. openlogos archive — close the proposal

The key insight: impact analysis evaluates the ripple effect across the entire artifact chain — requirements, design, API, DB, tests, code. No artifact falls out of sync. This is what separates structured iteration from “fix the code and forget the docs.”

logos/
├── logos.config.json # Project configuration
├── logos-project.yaml # AI collaboration index
├── resources/ # All phase artifacts
│ ├── prd/ # Requirements → Design → Technical plan
│ ├── api/ # OpenAPI YAML specs
│ ├── database/ # SQL DDL / schema
│ ├── test/ # Test case documents
│ ├── scenario/ # API orchestration tests
│ └── verify/ # Acceptance reports
├── changes/ # Delta change proposals + archive
├── skills/ # AI Skills (SKILL.md per skill)
└── spec/ # Methodology specifications

Full Project Structure →

OpenLogos is tool-agnostic. Built-in integration support for:

PlatformIntegration mechanism
Claude CodeNative .claude/ plugin system
OpenCodehooks/ and command integration
Cursor.cursor/rules/ auto-attached rules
Other toolsAGENTS.md as universal entry point
  • Quick Start — Install and create your first project in under 10 minutes
  • First AI Collaboration — Walk through a complete development session
  • Tour — See OpenLogos applied to real projects (FlowTask + Money-Log)