Kivo — Stop Guessing. Start Scoping. | Local-First AI for Freelancers

Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle Submission 💸

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

Ever had a project that starts with a clear list of requirements, but somewhere along the way, the client starts asking for “just one more thing”?

One small feature turns into another, and before you know it, you've spent hours working on things that were never part of the original agreement.

That's the problem I wanted to tackle with Kivo.

Kivo is a local-first AI tool designed to help freelancers and independent developers manage project scope without the awkwardness of constantly negotiating extra work.

One thing I really wanted to get right was the separation between AI and business logic. AI can help interpret a request, but pricing should be predictable, and a new requirement shouldn't become part of the agreed scope without the proper approval.

The goal isn't to make conversations with clients more complicated. It's to make expectations clearer for everyone.

You can explore the project, check out the screenshots, and find instructions for running it locally here:

🔗 GitHub: https://github.com/Txxjasss/Kivo

The screenshots use sample projects and illustrative data.

Kivo currently runs locally rather than as a publicly deployed SaaS.

The complete project is available on GitHub:

Feel free to explore the frontend, backend, tests, and documentation.

The AI handles the language-heavy part: understanding client briefs and identifying possible additions to a project. The backend handles pricing, persistence, and workflow rules.

I wanted the application to do more than simply wrap an AI prompt in a nice interface. That's why I also worked on baseline locking, approval workflows, duplicate-request protection, and automated tests.

The repository's test suite contains 23 automated tests across four test files, and the frontend production build has been verified.

There are still things I'd like to improve, but getting these different pieces to work together has been a valuable learning experience.

For a tool that deals with client briefs and project discussions, privacy matters.

Using the open-weight Qwen2.5 3B model through Ollama means Kivo can run inference locally without depending on a third-party cloud AI API key. It also gave me the freedom to experiment with the model and build the rest of the application around it.

I like that open innovation makes this kind of experimentation accessible. You can take existing tools, understand how they work, adapt them to a specific problem, and build something useful without having to rely entirely on a closed platform.

For Kivo, that freedom helped me explore an approach where AI assists with understanding requirements while the application keeps pricing and approvals under explicit control.

This is an early version, and there's plenty more I'd like to build, but I'm happy with what I've been able to put together.

I'm submitting Kivo for the overall Hacktoberfest Open-Source AI Challenge Week 2 prize. My focus is on building a useful product around open-weight AI, with local inference, transparent pricing, and a practical workflow for freelancers.