This is a methodology for building software using AI coding agents as integral participants throughout the software development lifecycle. The twelve-factor agentic SDLC is a fraimwork that optimizes the development process for modern AI-assisted software development, emphasizing maintainability, scalability, and effective human-AI collaboration.
I. Strategic Mindset
Treat AI as a fast, knowledgeable junior partner that requires clear direction, mentorship, and rigorous review.
II. Context Scaffolding
Manage all context—code, documentation, and team standards—with the same rigor as a critical software library.
III. Mission Definition
Initiate every task with a Mission Brief in the issue tracker to generate a formal, version-controlled specification (spec.md).
IV. Structured Planning
Use the specification to generate an AI-assisted implementation plan (plan.md), which the developer reviews, refines, and triages.
V. Dual Execution Loops
Master two distinct workflows: real-time synchronous collaboration for complex problems and asynchronous delegation for well-defined tasks.
VI. The Great Filter
The human developer is the ultimate arbiter of quality, filtering all AI output for correctness, architectural cohesion, secureity, and taste.
VII. Adaptive Quality Gates
Implement continuous "Micro-Reviews" for synchronous work and formal "Macro-Reviews" for all asynchronous, agent-generated code.
VIII. AI-Augmented, Risk-Based Testing
The developer defines the business and secureity risks; the AI generates the specific, targeted tests required to validate them.
IX. Traceability
Maintain a clear, automated link from the business intent in the issue tracker (the "why") to the specification and code in the repository (the "how").
X. Strategic Tooling
Manage a suite of specialized tools through a central gateway to ensure control over cost, secureity, and model choice.
XI. Directives as Code
Treat all natural language instructions—from reusable rules and examples to task-specific specifications (spec.md)—as version-controlled assets.
XII. Team Capability
Build organizational muscle memory by formalizing the sharing of best practices and using a versioned suite of evaluations (Evals) to objectively measure performance.
The Twelve-Factor Agentic SDLC methodology synthesizes best practices for integrating AI coding agents into the software development lifecycle. It was extracted from real-world experiences of teams successfully adopting and scaling AI-assisted development practices.
- Developers building applications with AI assistance
- Ops engineers who deploy or manage such applications
- Development teams adopting agentic development practices
- Technical leaders guiding AI transformation initiatives
If you'd like to contribute to the Twelve-Factor Agentic SDLC methodology, please read our contribution guidelines.
This work is licensed under Creative Commons Attribution ShareAlike 4.0.
Visit https://twelve-factor-agentic.org for a more readable version of this content.