How It Works

Zhenfy One uses a multi-agent pipeline where specialized AI agents handle each phase of the project lifecycle. Each agent is an expert in its domain.

The Pipeline

1. PM Agent (Brief)

The PM agent takes your idea and produces a structured project brief: goals, requirements, user stories, and success criteria. You review and approve before proceeding.

2. Tester Agent (Tests)

Writes a comprehensive test plan and test cases based on the brief. Tests are written before code, following test-driven development principles.

3. Security Agent (Security Plan)

Security is a default stage of every project. This agent builds a STRIDE threat model and an OWASP-mapped security test plan — including an executable suite that will attack the deployed app — and a reviewer hard-gates the plan before the build proceeds.

4. Ops Agent (Ops Plan)

Designs the operations plan for your project: the analytics dashboard, an uptime canary, and the alerts that fire when something looks wrong — each tailored to your project type — plus a runbook for ongoing operations. Hosting itself is handled by the platform.

5. Builder Agent (Build)

Writes the code based on the brief and test plan. The builder generates clean, production-ready code, and stays grounded in your brief — it uses only the content you provided and won't invent facts, names, or details you didn't ask for. For a full-stack app it first provisions an isolated per-project boundary (namespace, container registry, and datastores), builds the backend, and wires in AI features through Amazon Bedrock when the app needs them.

6. Test Runner

Executes all functional test cases against the deployed beta. If tests fail, the builder iterates until they pass.

7. Security Tests

Runs the security suite against the deployed beta, re-attacking it for issues like leaked data, broken access control, and injection. A failure sends the code back to the builder to fix before you ever see it.

8. Ops Implementation

Sets up the actual infrastructure according to the ops plan. Deploys the code to a beta environment for review.

9. Review

A final AI review checks everything end-to-end, then the pipeline stops at the beta preview. You review the beta and either request changes (which rebuilds) or promote it.

10. Deploy to Prod

Going live is a separate, explicit action: clicking Deploy to Prod promotes the reviewed beta to production with monitoring, alerting, and analytics. Reviewing the beta does not auto-deploy.