Frequently Asked Questions

Strategic answers regarding Security, IP, and Factory Logic.

1. Security & Intellectual Property

Q: Does Klyve train on my code?

A: No, Klyve does not train on your code. It runs locally on your machine and stores your specifications and source code in a local, encrypted database. It only sends specific code snippets to the LLM provider you select (e.g., OpenAI, Anthropic) for the purpose of executing the immediate task, subject to that provider’s API privacy policy.

Q: Can I use Klyve offline?

A: Klyve’s internal logic, file management, and database work completely offline. However, the core generative and development capabilities require an internet connection to reach your chosen LLM provider’s API.

Q: Does my code leave my machine?

A: Your source code files remain on your local machine. Klyve sends only the necessary context (prompts and relevant file snippets) to the LLM provider you configure (e.g., OpenAI, Anthropic, or Google) via their secure APIs. There is no Klyve backend on the cloud or anywhere else that receives your IP.

Q: Who owns the generated code?

A: You do. You own 100% of the code Klyve generates. It produces standard, readable code. If you stop using Klyve tomorrow, your project is just a standard codebase that can be opened in your IDE. There is no vendor lock-in on the output.

2. Technical Control & Quality

Q: What prevents Klyve from generating unmaintainable “spaghetti code”?

A: Unlike chat interfaces, Klyve is State-Aware.

  1. Standards: It strictly adheres to the Coding Standard document generated for your specific project.
  2. Review: Every block of code passes through automated syntax and style reviews before being committed.
  3. Context: It locally retains the context of existing architectural patterns to ensure consistency.

Q: What happens if the AI generates bugs?

A: Klyve uses a “Human-in-the-Loop” architecture. If the AI cannot solve a debugging problem within your configured limit, the system pauses and escalates to you. You can launch the Manual Fix workflow to open the code in your IDE, fix the logic, and resume the sprint.

Q: Can I change an approved spec document mid-sprint?

A: No. You cannot jump back and redo a document that is already approved. To change the scope, you must either edit the un-implemented backlog item or raise a new Change Request, which triggers a controlled spec update upon completion.

3. Operations & Cost

Q: Is there a monthly subscription?

A: No. The Klyve executable is free to download. Klyve follows a “Bring Your Own Key” (BYOK) model. You pay your preferred LLM provider directly for API usage, giving you complete control over costs.

Q: Which LLMs are supported?

A: Klyve supports several major hosted LLM service providers. You can also configure a Custom Endpoint to connect with local models or private cloud instances. You can additionally configure separate models for “Reasoning” and “Quick Tasks” to optimize costs.

Q: Can I use Klyve on an existing application?

A: Yes. Klyve includes the ability to intake and assess an existing application. When you point it to an existing directory, it scans your codebase, analyzes it, and prepares a set of reference specifications for it. This allows you to start managing changes to a legacy project with Klyve’s automated workflows immediately.

4. Regulatory Alignment & Compliance

Is Klyve compliant with the EU AI Act?

Klyve is designed for Compliance. Architected under the guidance of an IAPP Certified AI Governance Professional (AIGP), Klyve voluntarily adopts Article 14 (Human Oversight) standards to ensure you remain the Data Controller.

  • Human-in-the-Loop: The orchestrator halts for approval before every key action (Finalisation of Specifications and Standards, Planning, Coding, Testing).
  • Data Sovereignty: Your data resides on your local machine, and only limited project data is transmitted to your selected LLM provider. There are no Klyve servers.
  • Transparency: In addition to informatory pop-ups and notices to inform you about key AI actions, printed .xlsx files, .docx files and source code files generated by Klyve bear a watermarked footer to indicate that they are AI-generated.

Does Klyve support US Risk Management Standards (NIST)?

Yes. Klyve is architected to enforce the NIST AI Risk Management Framework (AI RMF).

Specifically, it addresses the Manage and Govern functions of the framework:

  • Map & Measure: Klyve forces a mapping of requirements before code generation begins. After artifact generation a requirements traceability matrix provides a trace of backlog items to generated artifacts to assure that none have been skipped.
  • Human-Centric Governance: By refusing to execute code without explicit human approval (the “Senior Partner” model) and the necessary project inputs, Klyve acts as a technical control against AI hallucinations and drift.

5. Factory Logic

Q: How does Klyve determine the implementation plan?

A: Klyve reads your selected Backlog Items and cross-references them with your Technical Specification. It uses a planning algorithm to determine the most logical sequence of changes (e.g., Database Schema → Backend Logic → API Layer) to minimize dependencies.

Q: What should I do if a Regression Test fails at the end of a sprint?

A: You have two strategic options:

  1. Immediate Fix: Pause, debug manually in your IDE, and retry the test.
  2. Acknowledge & Deferred Fix: Ship the sprint as “Completed,” but Klyve will automatically log a high-priority Bug Report in the backlog for the regression failure.

Q: How do I handle “State Drift” if I changed code outside of Klyve?

A: When you load a project, Klyve performs a Pre-flight Check (requires Git). If it detects uncommitted external changes, it asks you to commit them to ensure its internal understanding remains in sync with the file system.