Arclyne is a Windows-local AI research assistant for papers, lectures, web pages, and Zotero libraries. It accepts PDFs, arXiv papers, YouTube videos and playlists, web pages, and Zotero libraries in workspaces. Chats provide citations to a page, paragraph, or video timestamp, and the product says unsupported answers are flagged. PDF extraction and YouTube transcription happen locally; scanned PDFs are detected and flagged, while playlist videos process sequentially in the background. Local models run through a Llama.cpp sidecar on the computer’s CPU or GPU, with model matching during onboarding. Arclyne says sources are not uploaded, no backend account is needed, and Wi-Fi can be off; update checks are the only network call and can be disabled. The personal license is 49.00 USD once, covers two machines, and includes lifetime 1.x updates. A 14-day trial is offered. The Windows v0.9 download is listed as 84 MB.
Who it is for
It suits Windows users researching mixed local and online source materials who want cited workspace answers and offline operation. Hardware needs vary by model size, from 8 GB RAM for lightweight models to 32 GB or a dedicated GPU for larger ones.
What is good
- Works with PDFs, videos, web pages, and Zotero
- Citations link to source locations
- Sources stay on the machine
- Can work with Wi-Fi off
- One-time license covers two machines
What to know first
- Windows app; macOS release is pending notification
- Does not support GPT-4 or Claude
- Local models are less suited to frontier reasoning
- Scanned PDFs are flagged
Verdict
Arclyne focuses on private, local research across documents, videos, websites, and Zotero. Its one-time personal license and offline design are useful, while model and platform limits matter.
Arclyne plans and pricing
All plansCompared on AI research assistants
- Free plan
- No
- Source citations
- Yes
- Source access
- web+files+connectors
- File uploads
- Yes
