LoxiaFlow is a local-first AI-Driven Development Lifecycle (AIDLC) desktop app: AI assists at every phase from planning through testing and security, with plan-driven agents executing work when you want them to — while you set the scope, models, and permissions and validate the results. Every page below documents the exact fields, buttons, and AI behaviors you'll see in the app — not just a feature summary.
Everything runs on your machine — projects, documents, mockups, and workflows are stored locally, and there's no server or account required to use the app.
Installation →
Download and run the installer for your platform (access currently provided on request).
First Run →
What happens the first time you launch the app.
Core Concepts →
How a feature gets built end to end, from Style Guide and Project Overview to a testable feature.
AI Console →
Project-aware chat: the model writes SQL, LoxiaFlow runs it locally, only results go back.
How the agents work
LoxiaFlow runs an agentic loop on your machine. Plan: Workflows break each feature into ordered layers. Execute: agents run those layers, or queued issues in the Issue Runner, reading and writing files and running shell commands in your project. Verify: Success Criteria, the Test Generator, and Security Scans check the result. Iterate: you review changes in Git Changes and re-run issues that need more work.
An agent is the engine that acts in your project (Claude CLI, OpenCode CLI, or the LoxiaFlow Agent); a model is the LLM behind it. You choose both through Model Mapping. Queued Issue Runner runs skip permission prompts, and agents run one task at a time. LoxiaFlow is built for new projects; importing existing codebases is planned for a future release and isn't supported today.
Where to start
- New to LoxiaFlow? Read Core Concepts first — Layers and Workflows are referenced by almost every other page.
- Setting up a project? Go Project Layers → Documents → Mockups → Generate Workflows in that order.
- Once a project exists, jump straight to Git Changes, Test Generator, or Issues.
- Want to chat with your project? The AI Console lets the model query your project data on-device via generated SQL — nothing leaves your machine.