Your files and local storage
Updated October 3, 2026. This policy covers Cowslator’s document, image, audio, video and AI tools. Lochs Studios maintains the site; questions can be sent through Contact.
File processing
Normal chat workflows process your selected files on your device, without uploading them to a Cowslator processing server. An optional paired native Node can receive audio on your own device. The experience relay does not execute processing jobs.
Browser memory and deletion
Models, settings, extracted document chunks, embeddings and summaries use browser storage, including IndexedDB. Documents and retained jobs/files are encrypted when the Cortex vault is unlocked; otherwise retained data is local plaintext. Legacy plaintext data is not automatically migrated. Delete chats, saved documents or private memory using their controls, or clear all site data in your browser. Chat titles and transcripts are saved in IndexedDB and survive refreshes. Chats saved while Cortex is unlocked are encrypted; other chats are local plaintext. Session-only attachments and downloads require reattachment after refresh; retained jobs/files use the resume setting.
Cortex and integrations
Cortex stores bounded local workload measurements and deduplicates repeated tasks. It automatically shares signed, bounded processing experience with this site’s metadata relay: coarse hardware/backend classes, rounded aggregate timings/counts and evaluated runtime policies. File contents, filenames, prompts, embeddings, outputs and input/task/source hashes are excluded. A persistent public signing fingerprint accompanies this metadata. Processing and evaluation remain on your device; no computation is distributed. It does not upload your documents or train SmolLM2 weights automatically. Reviewed dataset exports contain the examples you select and require a separate export passphrase. External teacher connections need operator configuration and explicit release authorization. Optional MCP agent connections can request scoped tools on this device and exchange the same signed processing-experience metadata. Agents receive content-free job receipts; documents and downloads remain in the browser.
Website connections
The website downloads application code, runtimes and public models; some assets may come from third-party model/CDN hosts. Hosting providers receive connection data such as IP address, time and requested URL. Contact emails are handled by the sender’s and recipient’s email providers.
Analytics and advertising
The app uses Google Analytics (G-QS455YCMQ7); the site also uses AdSense (ca-pub-4324672100449694). These services may receive technical/interaction data and use cookies, identifiers, IP addresses or similar technologies for measurement and advertising, including personalization where permitted. See Google’s partner-site information, privacy policy and ad settings.
Your choices
Manage cookies and site storage in your browser and use the deletion and encryption controls in Options. Blocking storage or downloads may prevent tools from working. Available privacy rights depend on location. This policy may change as services change; its date identifies the current version.