sillon
App Store ↗
sillon
App Store ↗
Share a screen recording — a demo, a test session, a video call — and Sillon turns it into an illustrated brief: what was said, what was decided, and the screens that prove it. The video itself stays on your phone. A voice memo between two meetings feeds the same brief — the one for the right project, several memos at once, split with one gesture when they touch more than one. You approve line by line, and all of it lives in a markdown vault you own.
Ship the site redesign before September, V1 scope approved with the client.
A one-hour session shared into Sillon: the AI watches all of it, picks the screens that matter, crops them, timestamps them — and quotes the sentence that pointed at them. Each figure lands under its own step, and you approve it at triage. Here is the demo, in 38 seconds:
Four beats. You record and share, Sillon works while you get on with your day, and all you have to do is decide with a swipe.
You record your screen or talk into Voice Memos (or any app), then share it into Sillon — several captures at once. No new habit.
Sillon transcribes, works out the right project — existing or new — and prepares a proposal. You don't wait.
You approve or discard with one gesture. A memo that covers several projects? You split it right here — Sillon proposes the breakdown, you confirm. Nothing is written without your say-so.
Every approved capture merges into the project's brief — a synthesis that is up to date before your next check-in. A recorded session? The screens that matter land in it, cropped and timestamped, under the right step.
Four fields, always the same: Objective · Decisions · Open points · Next actions. Every approved voice memo merges into them — what is settled disappears, what is new is added, what changed is replaced.
Sillon re-reads the existing brief and proposes a merge: every sentence lands in the right field, duplicates disappear.
Two minutes before the meeting, you re-read one brief — not forty minutes of audio or twelve scattered notes.
Every merge is a git commit: you see what changed, when, and from which voice memo — and you can undo it.
Ship the site redesign before September, V1 scope approved with the client.
Sillon is not a closed box. Your notes are text files you own, versioned, and you can plug them into wherever you already work.
Every project is a markdown file in a git repository of your own — no proprietary format, no database you cannot export. The audio is deleted once the note is written: we keep the synthesis, not the raw material.
Our approach to data →Sillon writes to your vault first — that is the source. Alongside it, it can mirror every brief into your tools, optional and reversible.
Once Sillon is connected, there is nothing to learn: you ask your question in plain language — “what did we decide with Durand?” — and Claude, ChatGPT, Cursor or any MCP client goes and reads your briefs on its own. The AI is not digging through a raw archive: it reads clean briefs, already merged, that belong to you. And it is read-only: it consults, it never writes to your vault.
A real working document, as is: 22 citations, each one checked mechanically against the brief it points to. And whatever the briefs do not say is marked “hypothesis”, not asserted.
You never trust a model blindly. Sillon proposes, you approve — nothing is written without you. Every change is traceable and reversible thanks to the git history. Your notes do not decay over time: they get sharper.
Anyone can transcribe and summarize a memo. The difference plays out afterwards: what is left a week later, when you sit down to prepare your call?
Every memo becomes a clean note, filed next to the others. Synthesizing across notes is still your job: three months in, you have a well-written pile you never re-read.
Every meeting becomes one more note in their cloud, and the AI digs through them again at every question. The answer is good — then it evaporates. The consolidated document still has to be written.
Every approved capture merges into the right project's brief — Objective, Decisions, Open points, Next actions, and the screens that prove it, cropped and timestamped. The document exists before any question: up to date, versioned, in markdown you own, in Europe.
Based on their respective product pages and docs — August 2026.
The trial is free, no credit card, and only starts with your first approved brief: enough to merge about twenty captures across 2-3 projects and watch your briefs take shape on your real work.
No permanent free plan, no lifetime license: every memo triggers a transcription and then a merge by paid models — a real cost every single time. I would rather charge a straight price than fund the product with your data or with ads. Compare it to meeting intelligence tools (Otter, Granola…), not to note apps: Sillon hands you the deliverable, not one more note.
You talk into Voice Memos — the app you already use (Action button, Control Center, Siri, Watch) — then you share the memo into Sillon, one or several at once. No new habit; Sillon takes over in the background.
You share it like a memo. The AI watches the whole session, spots the screens that matter, crops them, timestamps them and quotes the sentence that pointed at them — each figure lands under the right step of the brief, and you approve the lot at triage. Only the audio and the cropped areas you approve ever travel: the video itself stays on your phone.
Four paths, and three of them start on your computer. If your video call produces a transcript (Teams, Meet, Zoom…), paste it into app.sillon.eu: it follows the same pipeline as a voice memo. If you recorded the session at your desk (QuickTime…), drop the file in the same place: it is handled like any screen recording. A YouTube video or a Claap recording? Just paste the link and the transcript is fetched for you (for Claap, with your workspace API key, set once in Settings). No transcript, no recording, no link? Put your iPhone on the desk and start a voice memo: the mic picks up the room, and you share the memo into Sillon after the call. Triage itself happens on your iPhone — that is where you approve.
In a markdown git vault that belongs to you, hosted in Europe. The audio is deleted once the note is written.
Yes — it is markdown, readable in Obsidian and in any editor; and Notion as an optional mirror if you connect your account.
No. The MCP connection is read-only: an AI can consult your briefs, never write to your vault.
Yes — by digging, at every question, through an archive of transcripts stored on their side. The answer is good, then it evaporates; the consolidated document still has to be written. Sillon takes the problem the other way round: every approved capture is merged into the right project's brief, a clean document that exists before any question — versioned, exportable, yours. The AI connected over MCP only reads it.
Because behind every memo there is a transcription and then a merge by a language model: a real cost every time, not once. A permanently free plan would mean throttling usage, selling your data or shutting down in six months — none of that holds for a tool you lean on before your progress reviews. That is exactly what the 14-day trial is for: test it on your real projects before paying. If Sillon does not save you a multiple of its price in prep time, do not buy it. Pricing is here.
No. What stops is the production of new briefs — the transcription and the merge, the part that costs money on every memo. What you have already written stays readable and exportable: in the app, in your git vault, and through your AI connection. If your vault is hosted with us, we keep it for 6 months after the subscription ends, then warn you 30 days before deleting it; if you use your own git repository, we never touch it. A tool you lean on before your calls has no business holding your notes hostage. The details are in the terms.
Yes — the iOS app is on the App Store, with a free trial and no credit card.
A 14-day trial, no credit card. Your recorded sessions and voice memos become living briefs, one per project, in a vault you own.