Show HN: Klyve - A local-first Software Factory (Automated SDLC) for solo devs

[mlinside]

Hi HN, I’m Mario. I retired in Dec 2024 after 35 years in software services and Operations Research.

I wanted to keep building, but I found the current set of chat LLMs frustrating for building complete applications. They are great at generating code (10% of the job) but not that useful when it comes to implementing specs for a full architecture, and then doing the testing and other project activities within the same context, and documentation (the 90% that makes the job a complete formal project).

I realized I didn't need a "coding partner". I needed a software services factory.

So, I built Klyve ( https://klyve.online ). It is a desktop orchestrator (Windows/Linux) designed to let Solo Senior Developers scale themselves by automating the SDLC, not just the typing.

The Core Philosophy: Determinism over probabilistic vibe

The Orchestrator (Not a Wrapper): Klyve treats the LLM as a stochastic component within a deterministic State Machine. It does not chat. It executes the SDLC as a strict workflow.

The "Senior Partner" Model: The AI is the junior engineer; you are the Lead. Klyve cannot execute code without an approved implementation plan, or any next step without approval, for that matter. It enforces a "Human-in-Command" workflow to keep it on track.

Full Lifecycle Management: It doesn't just write functions. It prepares and updates your documentation, generates unit tests, runs regression suites, and helps you manage a backlog.

Privacy & Architecture

I’m old school. I don't want my IP on a cloud server. Klyve is local-first and BYOK (Bring Your Own Key). It uses an encrypted local database for project state and only sends specific task contexts to the API.

Why I'm releasing it:

It is in Beta and free to download. It does work, but I have no plans to build a startup. I developed Klyve because I wanted to prove my own hypothesis that implementing a full orchestrator pattern would be a good way to build serious software with current models, although some limitations still remain.

I’d love feedback on the workflow logic.

( Note: For those interested in the governance aspect, I wrote another post on LinkedIn about how I used the EU AI Act to add the "Human-in-the-Loop" controls for this tool: https://www.linkedin.com/feed/update/urn:li:share:7419287932596936704/ )

Mario Lewis IAPP AIGP