Privacy & your data
nenva is private by default. No remote database, no account. Data is stored on your Mac in the standard macOS Application Support area. AI runs where you choose: your Mac (default), a server you own, OpenAI/Anthropic with your own key, or nenva Cloud if you chose it. Backups and storage location live in Settings → Storage & Backup. Transparency logs of every AI call and web search are in Settings → AI Assistant → Logs, machine-local, never synced.
Sync
Sync. Off by default: your data stays on your Mac until you turn sync on (setup asks, or Settings → Storage & Backup → Sync Between Macs). When enabled, changes travel between your Macs through your own iCloud Drive, encrypted. Merging happens on app start or refresh (Cmd+R), never mid-work, and the titlebar briefly shows "Synced N changes". Machine-specific things (email cache, model choices, API keys) deliberately don't sync.
Getting around
The sidebar down the left lists every app, grouped by what it is for, and marks the one you are in. Home is its first row. Collapse at the bottom narrows it to icons when you want the room back; it stays that way until you expand it, on that Mac only. More apps beside it opens the same show/hide list as Settings → Appearance; an app you hide leaves the sidebar but stays findable by name in Cmd+K.
The titlebar carries the rest: the nenva wordmark is the way home from anywhere, the grid button beside it opens the same list as a quick popover, and the button on the far left goes back to where you came from, naming it.
Keyboard shortcuts
- Cmd+K: the launcher and search in one. It opens with your most-used apps listed; type to narrow to any app, or to a note, task, project, journal entry or bookmark by name.
- Cmd+/: open the assistant panel over whatever you are looking at
- Cmd+R: refresh (also pulls sync changes from other Macs)
- Esc: close the open post, menu, or overlay
- Enter in any quick-add box: create the item
Documents and your writing voice
Documents: the user's repository for their digital documents (PDFs, scans and photos of paper read with the Mac's own OCR, spreadsheets, Word/PowerPoint files, RTF, web pages, notes), parsed and indexed on your Mac and ORGANIZED WITH TAGS. Import files, drag-drop anywhere on the page, or Finder-drop into the folder (~/Anjadhe/library/); supported: .pdf, .png/.jpg/.heic and other images, .docx/.doc/.rtf/.odt, .xlsx/.csv, .pptx, .md/.txt/.html. Tags: hover a row → "+ tag", or the tag strip while reading; a slash makes a level ("Finance/Taxes/2025" shows under Finance, Taxes and 2025 in the tree on the left); click a chip to filter, × to remove, "Rename tag" renames it everywhere; to create a child tag, hover a tag in the tree and click its "+" ("+ New" at the top for a top-level one), name it, and it is selected and ready: drop files on it (or anywhere while it is selected) to import them tagged, drag a document row onto it, or hover a row and click "+ tag" (the picker starts with the selected tag's path); "Untagged" collects what still needs filing. Searching shows passages grouped by document (inside the selected tag when one is chosen); opening one shows the text, TIDIED WITH AI once per document (paragraphs rejoined, headings/tables restored, nothing summarized or added; saved once as a Markdown sidecar in the folder's hidden .anjadhe/tidy so it is never regenerated for an unchanged file, on any Mac sharing the folder; automatic when the AI model is on your Mac, otherwise a "Tidy with AI" button unless Files may leave this Mac in Cloud privacy), or, when the AI model can SEE images (a local model with its vision file, or a vision-capable cloud model), READ FROM THE PAGE IMAGES by the vision model (far more accurate than the PDF text layer; that transcription becomes the document's text for search and the assistant; first 25 pages automatic, the rest on a click; a document parsed or OCR'd earlier shows "Re-read with vision" in the banner to reprocess it), with a banner saying which view is up and "Show extracted text" to flip back (matches are marked in the extracted view), plus Open original / Show in Finder / Ask about this document (the assistant then knows what's open; it always reads the extracted text, never the rewrite). The assistant can search (search_library, optionally within a tag), list by tag (list_documents), read a document (read_library_doc) and tag documents on request (tag_document). Cmd+K finds documents by title or tag. Deleting a document moves the file to the macOS Trash. Semantic search needs a one-time ~330 MB model download in Settings → Documents (also index status, Rescan/Re-index, the folder location); until then search is keyword-only. Parsing, OCR and indexing always run locally; only passages retrieved for a question travel with that chat turn. Tags sync between Macs (the index is rebuilt on each).
Writing voice: the assistant can learn how you write and draft in your voice WHEN ASKED. It is a setting, not an app: Settings → Writing voice. Turn it on and pick what it learns from: documents you add (Add documents… or drag-drop onto the page), your Notes, your Journal, and your sent emails (only the parts you wrote: quoted replies, forwarded messages and signatures are stripped before anything is studied). Each source is a switch you can change any time; nothing is read until you press Study. Study reads a spread sample on the AI model you chose (Study depth on the same page sets how much) and writes a short, editable style guide plus verbatim passages of your writing. Edits to the guide stick (re-studies never overwrite them) and passages can be pinned or removed. To use it, ask any chat: "write this in my voice", "make it sound like me", "reply the way I would". The assistant fetches the guide and passages first, then writes; Draft in my voice… on the settings page opens a chat already primed. There is one voice, your own; voices built from other people's writing are not a feature. Routines and other automatic writing use the assistant's own voice. Turn off removes the guide and passages; documents, notes, journal and emails stay where they are. No fine-tuning, no hidden style profile: the guide is a page you can read and change. Documents given to the voice are also visible in the Documents app, same folder.
Build apps with a coding agent
Prefer a terminal? Apps of your own are built with a coding agent, like Claude Code. They live in a plain folder on your Mac and nenva loads them automatically, right next to the built-in apps.
Where apps live. Turn on Build Apps in Settings. nenva creates ~/Anjadhe/apps/ in your home directory:
~/Anjadhe/apps/: one subfolder per app. Each has a manifest, the app's code, and its own saved data.~/Anjadhe/apps/CLAUDE.mdandAGENTS.md: the full contract for building an app (the manifest format, theAnjadheSDK for storage, navigation and tools, and worked examples). Both files hold the same instructions; a coding agent reads whichever one it looks for.~/Anjadhe/apps/.anjadhe-schemas.json: describes the built-in data shapes (notes, goals, schedule, and so on) so an app can read them.
Point a coding agent at it. Start your coding agent in that folder and describe the app; the agent picks up the contract on its own:
cd ~/Anjadhe/apps
claude # or whatever launches your coding agent
Because CLAUDE.md / AGENTS.md sit in that folder, the agent finds the whole contract by itself. Then just describe the app you want. For example:
Build a reading tracker app for nenva. I want to add books with a title, author, and status (want to read / reading / finished), see them grouped by status, and mark one finished. Follow the manifest and SDK in CLAUDE.md, and save data with the nenva storage API.
It updates live. nenva watches the folder: when the agent writes or changes a file, your app reloads on its own, no restart. A new app shows up in the sidebar and in Cmd+K next to the built-in ones.
If an app has a problem, nenva writes the error to .errors.log inside that app's folder. Point your agent at that file and it can read the error and fix it.