A nervous system for your company.
Sparrow runs on each person's own machine. It learns their work as it happens, notices blockers early, and tells whoever can clear them exactly what they need to know. The raw record stays where it was read, which is why it can be trusted to know so much.
Every hire adds a channel to everyone already here.
Five people share ten lines of communication. Seven share twenty-one. Standups, status docs, and check-in meetings exist to keep those lines from going quiet. With Sparrow the lines stay live on their own, so the meetings can go back to being about the work.
channels = n(n−1)/2 · with Sparrow, one each
Sparrow holds every channel. Each thread between two people runs through their Sparrows, so a person only picks one up when there is something on it worth their attention.
It reads what you let it see, then tells the one person who needs to know.
Sparrow lives on each person's laptop and picks up their work as it happens. When something matters to someone else, it drafts the message, you approve it, and that's all that leaves.
- Zoom, eng meetingLogin feature slips two weeks.
- Task boardNew due date: Oct 3.
- Slack, #marketingLaunch post goes out Tuesday, 9am.
- Google DocThe post: “Introducing one-click login.” Owner: Dev.
- CalendarDev out of office Monday.
- Sparrow’s memoryDev, in a meeting transcript: “Slack me, I never see email.”
How it works
A map of your work
Sparrow reads what you let it see: meetings, your screen, the tools you already use. From that it keeps a map of your work and who is in it. The map stays on your laptop.
Who needs to know
It works out when someone else needs to know something, who that is, and exactly what they need in order to do their part.
Bigger jobs, minus the sensitive parts
For a task that needs a bigger model, Sparrow sends just that task, with names, numbers, and recordings removed first. The model does the work. Your Sparrow keeps the picture.
A nervous system for the company.
Most of a manager's week is relaying: direction downward, status upward, blockers sideways. Sparrow does the relaying, so the people can do the managing.
Set direction once, and hear about the things you'd never have had time to go looking for.
See what is moving, what is stuck, and who is waiting on whom, without asking.
Interrupted only for what you can act on. The quick question gets answered by your Sparrow.
Your systems can report in too.
Your GPU queue, your CI, your nightly jobs. We are building Sparrow to read them the way it reads a meeting, through an integration, an API, or the screen itself, so that when a job fails at 2 a.m., only the one person who can fix it hears about it. Design partners with a shared cluster will shape this first.
Let a local AI know everything about your work. Let no company know it.
Sparrow learns from the raw record on your laptop, so it can know more about you than any cloud assistant safely could. When it needs outside help, it sends one redacted slice: no names, no numbers, no recordings, just the context the question needs.
“Priya said Acme’s renewal slips if login isn’t done by Oct 3. That’s $240k.”
Highlighted parts never leave.“A two-week delay on a login feature may affect a customer renewal deadline. Draft a one-line note to the account owner.”
Name, company, date, and amount removed.Nothing to breach
Transcripts, screens, and notes stay on the laptop they came from. There is no central copy to leak, subpoena, or sell. There is no copy for us to read, and none for your employer either.
Built to run small
A model distilled for one job, understanding one person's work, costs a fraction of a frontier model per token. We are training that model now. Today the reasoning runs through a frontier model, sent only a redacted slice of the record, under a cap you set. The recordings never go.
A person approves every action
Sending, buying, or changing anything stops for a person to approve, and what you approve is exactly what runs. Something the model read or heard can inform a draft, and that is as far as it goes.
Delete it all, and get a receipt
Stop capture in one click. Delete everything in one more, and get a receipt naming every directory that was emptied. A deleted memory is gone everywhere it ever existed, because it only ever existed in one place.
Your Sparrow
Recording, transcription, who-said-what, memory, and search all run here, on an ordinary CPU. No GPU, no server.
A frontier model, working for your Sparrow
It gets one redacted slice, filtered by your rules, under a spend cap you set. Never the recordings. As the small model we're training gets better, more of this moves back onto the laptop.
We are taking three teams this fall.
Small teams doing real work with real deadlines, ideally with something that already coordinates badly: a shared cluster, a distributed team, a manager stretched across too many threads.
What you get
- A founder on every install call.
- A direct line, answered the same day.
- Fifteen minutes a week, and a readout at the end.
- A say in what gets built next.
What we ask
- Five to fifteen people who will use it on real workdays.
- Volunteers only. Ambient capture is a personal decision.
- The honest answer when it stops being useful.
What we promise
- Your data never reaches us. We built no way to receive it.
- Stop in one click. Delete in one more, with a receipt.
- Leave whenever you want. No explanation needed.
A working prototype, shaped by the teams who use it first.
Sparrow runs today on our own machines. So far that's the only place it has run, which is what the design partner program is for. Windows first, macOS next.
Capture from meetings and the screen. A searchable memory where every fact links to the moment it came from. A graph of people, commitments, and open loops. Two assistants paired so one can ask the other, under each person's disclosure rules. Actions in the browser and on the desktop that stop for approval. One-click stop, export, and delete with a receipt.
The distilled model that moves the reasoning onto the device. Integrations with the machines a team depends on. The flow of priorities down and status up across an organization. These are the decisions design partners get a say in.
Privacy Policy
This policy describes how Ravenry (“we,” “us”) handles information when you visit ravenry.us and when you use Sparrow, our on-device team-sync product. It also explains how Sparrow accesses, uses, stores, and shares Google user data when you connect a Google account.
Who we are
Sparrow is made by Ravenry. Sparrow runs on each person's computer. Our product promise is that Ravenry does not operate a server that receives or holds your team's work data. Questions about this policy: contact@ravenry.us.
This website
The marketing site is a static page. We do not run analytics cookies, advertising trackers, or an account system on the site.
- Design-partner form. Submitting the form opens your email app with a draft to contact@ravenry.us. Nothing you type is stored by this website. If you send the email, we receive whatever you chose to include (name, work email, company, team size, and your message) so we can reply about the design-partner program.
- Fonts. The site loads fonts from Google Fonts. Your browser may contact Google's servers to fetch those files. That request is governed by Google's policies, not by Ravenry.
- Hosting. The site is hosted on Cloudflare. Cloudflare may process standard technical request data (such as IP address and user agent) to deliver the page and protect the service.
Sparrow on your device
When you install Sparrow, capture, memory, and reasoning stay on your machine unless you approve something leaving. You control what Sparrow may see (for example meetings, screen content, or connected tools). You can stop capture, export your data, and delete it, including a deletion receipt where the product provides one.
Ravenry does not receive your on-device work graph, transcripts, or screen content through Sparrow's normal operation. We have not built a path for that data to reach us.
Google user data
If you connect Google (for example Google Docs, Google Calendar, or other Google Workspace services Sparrow supports), Sparrow requests access through Google's OAuth consent screen. We request only the scopes needed for features you use.
- Access. With your authorization, Sparrow reads Google user data you choose to connect—such as document contents or metadata, calendar events, and related account identifiers—so it can notice coordination-relevant facts on your device.
- Use. Google user data is used only to provide and improve user-facing Sparrow features that are visible in the product: local memory and search, drafting messages for your approval, and syncing status with teammates under rules you set. We do not use Google user data for advertising, credit decisions, or sale to data brokers.
- Storage. Google user data obtained through Sparrow is stored on your device (and, where the product offers it, in exports you create). Ravenry does not store your Google user data on Ravenry servers.
- Sharing. Google user data does not leave your device except when you approve an outbound action (for example a redacted message to a teammate's Sparrow, or a capped task to a frontier model). We do not transfer Google user data to third parties for their independent use.
- Revocation. You can disconnect Google access in Sparrow and in your Google Account permissions. After disconnect, Sparrow stops new access; you can delete locally stored copies as the product allows.
Google API Services User Data Policy — Limited Use
Ravenry's use of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements.
- We use Google user data only to provide or improve user-facing features that are prominent in Sparrow.
- We do not transfer Google user data to third parties except as needed to provide those features with your consent, for security, to comply with law, or as part of a merger or acquisition with your prior consent where required.
- We do not allow humans to read Google user data unless you give affirmative agreement for a specific case, it is necessary for security or to comply with law, or the data is aggregated for internal operations in line with applicable law.
- We do not sell Google user data, use it for ads (including retargeting or personalized advertising), or use it to determine credit-worthiness.
Children
Sparrow and this site are not directed to children under 13. We do not knowingly collect personal information from children.
Changes
If we change how we handle Google user data or other personal information in a material way, we will update this policy and, where required, ask for fresh consent before using data in a new way.
Contact
Ravenry · contact@ravenry.us · https://ravenry.us