Make something
of what you know.

Tenvar is a private desktop workspace for research, transcription, notes and writing. It brings documents, recordings and imported notes into projects, uses on-device AI for transcription and cited questions, and turns selected evidence into notes, manuscripts and reports. It suits researchers, writers, students and people working with interviews, reference documents or client material who want to keep that work on their own computer.

Tenvar focuses on your own local sources: importing them, reading or transcribing them, asking cited questions, marking evidence and writing from it. It uses models installed on your machine rather than sending prompts and documents to a cloud AI provider. It is not a replacement for a general web-search assistant or a guarantee of the same capabilities as a hosted model.

Core reading, transcription, OCR, search, questions, writing and export work locally once the required models and components are installed. Model or document-runtime downloads, optional online web/email import, requested update checks, and paid-license activation or revalidation need a connection. Your source library is not uploaded for AI processing.

No account, credit card or cloud AI API key is needed to start using the Free plan. Models run on your own hardware. Buying a paid license is a separate purchase through Polar; the license key is then activated in the app.

Free includes one Tenvar project plus one imported library: choose either an Obsidian vault or Apple Notes. The imported-library allowance is shared, not one of each. Files within those allowances and the available tools are unlimited, with no time limit, transcription-minute quota, export watermark or feature paywall. Apple Notes import is Mac-only.

The $59 USD early license removes the project and imported-library limits. It is a one-time purchase that includes version 1 and its updates; the planned version-1 price is $129. It is not a subscription. Activation and periodic validation check the license; being offline does not disable an already activated license.

A dedicated graphics card is not required. An 8 GB computer can use smaller models, and Tenvar checks memory and hardware when offering models. Larger speech, language and vision models need more RAM or VRAM and storage; some are unsuitable for smaller machines. Speed varies. Standard Windows and Linux builds support CPU and compatible Vulkan graphics; optional NVIDIA builds add CUDA. Apple Silicon uses Metal where supported.

No. New projects start unencrypted. Choose Encrypt project and set a project password or PIN to protect that project. This secret is separate from your computer login, and Tenvar cannot recover it if forgotten. Original files outside Tenvar are not encrypted by importing them. App access locking and project encryption are different protections.

No. You can browse a vault through Shelves and bring selected files into a project, or convert it into an editable Tenvar copy. The original vault is left alone. Later edits are not written back, and Obsidian plugin behavior or every formatting convention is not guaranteed to carry over.

On macOS, choose individual notes from Shelves or import an Apple Notes library. The app requests the system permission needed to read Notes and creates local copies. Rescan can bring in updates while preserving conflicting Tenvar edits. Text and available attachments have fidelity limits; locked or unreadable notes are skipped and problems are reported. It is not a full-fidelity Apple Notes backup or two-way sync.

Yes. A self-contained HTML report can include selected playable audio and video evidence and can be opened in a browser offline without installing Tenvar. PDF is available for a printable handoff, but it does not preserve interactive media playback.

No. Models can misunderstand speech, assign a speaker incorrectly or produce an inaccurate answer. Review the transcript and speaker assignments, and open citations to compare answers with the original. Language coverage, OCR quality and output quality depend on the model and source material.

Tenvar is early software and can have rough edges. There is no hosted collaboration or cloud library sync, no two-way Obsidian or Apple Notes sync, no live meeting captions, and no system-wide dictate-anywhere tool. Continuous source-book read-aloud and report-wide AI editing are not in this release. Hardware, import fidelity and document runtimes impose additional practical limits.