SpecViber stores the metadata it needs to coordinate the work — titles and paths, states, checklist lines, relations, tasks, comments and history. Your markdown text and your code stay in your project.
That distinction matters for consultants, agencies, and teams working under confidentiality, NDAs, or data-protection obligations. The rest of this page is deliberately precise: where details are public today, they stand clearly; where they haven't been published yet, they're marked as such.
SpecViber gives the developer and the AI agent one shared working layer. The shared layer holds metadata that coordinates the work — document titles and paths, properties, the text and state of checklist lines, relations, assignments, history, tasks and comments. The markdown text and source code themselves remain on the developer's machine and in the developer's repository.
The agent reads and writes the shared layer through MCP tools, instructions, and skills. It does not require uploading project source code to a SpecViber-hosted service.
The practical boundary is straightforward. SpecViber does not need the project itself to move into a hosted workspace. It coordinates the work by storing metadata around the project, not by taking ownership of the project files.
That boundary is what makes the model easier to evaluate for consultants and confidentiality-bound teams.
Hosting, retention and sub-processors are described in the Privacy Policy, and the service terms in the Terms of Service. Pricing-related deployment terms, certifications and exact integration scope are not yet published. If you need these for evaluation, ask directly.
If you're reviewing SpecViber for professional use and need exact answers about deployment, data handling, repo access, or privacy — contact us and ask for current status.