1.3.0

🎁 Azynote is now free#

No trial, no license, no tiers, no account. Download it and use it, every feature included.

Here is the honest reason. Running entirely on your own machine was always the goal of this project. Transcription and storage have been local since day one, but the intelligence, the part that turns a transcript into something you actually use, still ran in the cloud. So I was charging for a local-first promise while the most interesting step still left your Mac. Local models are finally good enough to be the default (see below), and not yet good enough that I would ask you to pay for that promise. Charging for a half-finished promise is the wrong trade, so I am not charging at all. Details on the Pricing page.

🖥️ The AI now runs on your Mac#

Azynote ships with a local model it downloads and manages for you. It powers OnePagers, chat, the Command Panel, and search indexing, all on-device. With it selected, nothing about your meetings leaves your machine at any point. It needs 16 GB of memory or more, and about 6.8 GB of disk for the two models it downloads: one that writes, one that searches. If a model file ever goes missing, Azynote notices at startup and offers to fetch it again.

🔍 Search runs on your Mac too#

Semantic search across your OnePagers used to need a Gemini key, even when everything else ran locally. Azynote now ships a second, smaller model built for search and downloads it during setup, so indexing and semantic search work with no cloud provider at all. Pick OpenAI or the Claude CLI for the writing and search still stays on-device. Index settings now tells you what is actually missing, instead of pointing everyone at Gemini.

🎛️ Choose your engine on first launch#

Opening a workspace now starts with a Choose your AI engine screen: Local, Gemini, any OpenAI-compatible provider, or the Claude CLI. Local is pre-selected, unless your Mac has under 16 GB, in which case Gemini is. You can switch any time in Settings > AI Configuration, and there is no "skip", because a choice you never made is how people end up wondering why nothing works.

📏 The context window fits your Mac#

Instead of a fixed window, Azynote now sizes the local model's context from the memory your machine actually has, measured against real KV-cache cost rather than a guess. A 16 GB Mac gets a solid working window, a 64 GB Mac gets a very large one, from the same model. Long sessions that used to overflow now either fit or tell you clearly that they do not.

⚡ A faster first prompt#

The chat model pre-warms when you open chat, with a visible cold-start indicator, and the Command Panel no longer freezes the UI while the model wakes up.

🧰 While you are here: it doesn't stop when the meeting ends#

None of this is new in 1.3.0, but it is easy to miss if you only ever open Azynote for a call. Smart Actions (Ctrl+Shift+O) transform text you have selected in any app on your Mac: proofread, improve, translate, summarize, or any action you define yourself. Quick Capture (Ctrl+Shift+K) drops a thought into your workspace from wherever you are, without breaking what you were doing. And chat (Ctrl+Shift+C) works across your whole workspace, not just meeting history, including any MCP server you connect to it.

🐛 Notable fixes#

Happy meetings! 🎙️