A new context layer
A context relay is a live channel between your work and your agent. Whatever conversation you're in, whatever you just said, it's already there, so you don't have to explain it again by hand.
00 / What a relay carries
Cyrano is an audio-driven relay: what's said around you is read and classified where it happens, and it arrives in the one format your assistant already works in.
The tool you use doesn't change. What it can take in does. And the same rail carries anything else you'd otherwise move by hand, which is why the relay still earns its keep on days you're not on a single call.
Speech is the richest thing in a working day and the one your tools never see. The same rail carries the window in front of you, a pasted screenshot, or the passage you just copied. All of it on demand: keep a session rolling through a whole call, or catch a single burst when something lands.
Each signal is read on your device and reduced to text, then handed onward on the same terms as speech. You press something, it goes, and nothing else does.
What did we just commit to?
Beta ships Thursday, and the pricing page is yours by Wednesday.
Nothing to select, paste, or explain. The question you wanted to ask in the first place is the only thing you type.
01 / Closing the gap
Your assistant can reason over almost anything you give it. The problem is the giving.
A useful exchange happens in a call, at a desk, or while walking between rooms. Later, you select the relevant passage, copy it, switch applications, add a caption, and ask the question you wanted to ask in the first place.
The relay removes that transfer step, without asking you to build a new habit. It just takes the friction out of the one you already have.
You paste notes into an assistant when a conversation matters. The relay makes the useful context available before you remember to move it.
You already do this every day. The relay removes the repeated selection, labeling, and handoff that have become invisible overhead.
You start every session. The relay gives your agent a bounded, current view only when you choose to capture and connect it.
02 / The keystroke argument
A short conversation can contain hundreds of words. Moving it into an agent shouldn't require hundreds of keystrokes and a six-step detour.
The real saving isn't the typing. It's the attention you spend deciding what to select, where to put it, and how to explain what it is, an interruption that repeats every time the physical world needs to meet the model.
6 transfer actions · hundreds of characters moved by hand
0 copy-paste actions · 0 transfer keystrokes
Next
See the work it supports, and the boundary that keeps augmentation honest.