From Idea to API in 2 Days: Building Bankly with Agentic Workflows

Bankly, a backend banking system built in just two days using Apollo GraphQL and Prisma. Designed as a self-technical test and proof-of-concept that showcases how agentic workflows and AI-assisted development can accelerate modern software delivery while maintaining quality and structure.

Here's the complete commit-by-commit progression showing how AWP enabled systematic development without losing momentum:

Vibecoding with Agentic Workflow means:

• Working in a flow state of high creativity and speed
• Letting AI co-pilots handle tasks within a structured framework
• Avoiding cognitive burnout or "over-vibing" through systematic boundaries
• Maintaining clarity and context throughout the development process

The job market is evolving, and engineers must adapt. AI is no longer optional—it's a multiplier. But scaling engineering with AI requires maturity, structure, and awareness of risks.

I call this risk "over-vibing"—a state where you push too far, too fast, and lose context. AI tools (especially LLMs) suffer from context drift. So do humans. When you're in the flow but lose track of the bigger picture, productivity can quickly turn into technical debt. To manage this risk, I created a "AWP" that maintains structure while preserving creative momentum.

The secret behind Bankly's rapid development lies in a methodology I call "Vibecoding with Agentic Workflow". This approach combines high-creativity flow states with structured AI collaboration, avoiding the pitfalls of uncontrolled automation or "over-vibing." The proof is in the commits: Bankly was built on June 24-25, 2025 in two days with 35+ structured commits following the Agentic Workflow Protocol (AWP). Each commit represents a deliberate step in the development process, from initial setup to pre-production features. It means we can scale up the development and reduce the risks like loosing the context, getting lost in complexity, or limits for both Humand and AI.

---
header: AWP - Agentic Workflow Protocol
legend:
  - color: "#3B82F6"
    text: "Planning Stage"
  - color: "#6366F1"
    text: "Development Process"
  - color: "#8B5CF6"
    text: "Documentation Flow"
  - color: "#10B981"
    text: "Version Control"
---
graph TD
    classDef planning stroke:#3B82F6,stroke-width:3px;
    classDef development stroke:#6366F1,stroke-width:3px;
    classDef documentation stroke:#8B5CF6,stroke-width:3px;
    classDef versionControl stroke:#10B981,stroke-width:3px;

    subgraph Planning
        A[Define Tasks] --> B[Iterate Tasks]
    end
    
    subgraph Development
        C[Review Scope] --> D[Development]
        D --> E[Check Context]
        E --> F[Run Tests]
    end
    
    subgraph Documentation
        G[Documentation] --> H[Final Check]
    end
    
    subgraph Version Control
        I[Git Review] --> J[Next Task]
    end
    
    B --> C
    F --> G
    H --> I
    J --> B
    E --> C

    class A,B planning
    class C,D,E,F development
    class G,H documentation
    class I,J versionControl

MCP Agentic SLDC (Software Lifecycle Development Cycle) is an open-source protocol designed to help teams and AI agents maintain clarity, transparency, and flow throughout development.

Key Benefits:

• Always know the next step
• Restore context after breaks or resets
• Keep documentation and code in sync
• Balance human-AI collaboration effectively

• VIBE-CODING: Productive creative flow in development
• OVER-VIBING: Losing control of the project due to excessive automation or drift
• Agentic Collaboration: Assigning structured roles and ownership to humans and AIs

Here's a simplified diagram of the development loop:

And how you regain context after a break:

Since we are in early stage, every project using Agentic SLDC with Agentic Workflow Protocol should include a file structured like this example:

The YAML also defines procedures like update, commit, next, and rules for AI-human handoffs.

As Vibe-Coding is new, as well as all the concepts, you are sill in full control, may you want it or not. Consequently, keep that list in mind:

• Use conventional commits tied to each step
• Keep README.md and SLDC YAML always updated
• Define ownership: Human vs AI
• Allow structured handoff between agents
• Recover from memory loss with a single check command
• Maintain modular, testable code architecture
• Document decisions and trade-offs as you go

Bankly was completed in 2 days because:

• The roadmap was clear and structured via AWP protocol
• Work was modular and agentic with defined ownership
• AI handled structured tasks like schema, model generation, and setup
• Human focus stayed on architecture, logic, and review
• Context was never lost thanks to the SLDC protocol
• 35+ commits show systematic progression without feature creep
• Day 1 focused on core functionality, Day 2 on production readiness

This isn't just productivity—it's sustainable velocity. The system prevents burnout while maintaining quality and allowing for rapid iteration.

MCP Agentic SLDC or my AWP - Agentic Workflow Protocol - is not just a tool—it's a philosophy for modern development:

• Build faster, but stay grounded
• Collaborate with AI, but maintain strategy and ownership
• Use clarity and process to avoid burnout and confusion
• Structure is the new speed in an AI-assisted world

Whether you're Vibe-Coding solo or managing a team of agents, structured collaboration is the key to sustainable, high-quality software delivery.

Contribute to MCP Agentic SLDC and join the movement toward human-centered, AI-supported development -> here.