Vinyasa Privacy & Data Security
Understand Vinyasa browser storage, tab-only AI credentials, consent-based provider requests, remote MCP processing, and local project backups.
Introduction
Privacy & Data Security
At RV Anveshana, we understand that your designs and mockups are intellectual property and often contain sensitive information. Vinyasa is built with a Privacy-First and Local-First architecture.
Local-First Architecture
Ordinary canvas editing and project autosave happen in your browser. Projects use IndexedDB; theme preference uses localStorage. Vinyasa does not automatically synchronize projects to a cloud account. Browser storage is not a backup: clearing site data, private browsing, or storage eviction can remove projects. Export .rvv files through File > Export As for durable backups. Files may contain design text, image references, and other sensitive content. See YAML and Git backups.
Data Processing
Local-first does not mean no network traffic. Loading the website, fonts, or remote images contacts their hosts. AI generation contacts the configured model endpoint only after consent. Using a remote MCP endpoint sends the document content supplied by your MCP client to its host for processing. Hosting and provider logs are governed by their respective policies; local storage does not protect a device from malicious extensions, shared-device access, or compromised software.
Working with AI
Built-in Copilot sends your prompt and, in Modify draft mode, a document summary plus requested element/schema batches to your chosen provider. Replace document mode starts with a blank document. Consent is required before generation; inspect your draft for secrets and personal data first. API keys and configuration are kept in tab memory, not written to localStorage or exported projects. Keys are sent to the selected provider for authentication. The provider's retention, training, billing, and regional-processing rules still apply. Cancel stops further requests but cannot recall data already transmitted. External ChatGPT, Claude, or coding-agent workflows follow the policies of the tools you use. Read AI setup, local models, and MCP security.
Our Commitment
Use synthetic records for prototypes, export backups regularly, close tabs on shared machines, and choose provider servers appropriate for your organization's data rules. Use hosted Studio at /editor and hosted MCP at /mcp. MCP tools are stateless and do not persist projects, but the MCP client and hosting platform may maintain their own records. Browser project storage does not mean Vinyasa is a locally installed application or that hosted MCP processing happens on your device.