More passion projects than hours
Sunny Luthra's launch post is two paragraphs of problem before a single word of product, and it reads like a diary entry: 'I have more passion projects than hours, and once coding agents made it possible to build them all at once, the work started scattering: plans in scrollback, to-dos in my head, every project pulling me away from the last. snyvi is what I built for that.'
That scattering is the unspoken tax of the coding-agent era. Agents made it cheap to start — a weekend project is now a Saturday afternoon. But every project you start this way generates plans that live in terminal scrollback, decisions that live in your head, and context that evaporates the moment you switch windows. The bottleneck moved from writing code to keeping track of five things at once. Sunny, who runs the product engineering studio mRova — a shop that builds and runs its own products, from a dog-boarding platform to a supplement brand with 5,000+ orders — felt it as an operator, not a theorist. So he built the tool he needed: one desk per project, the agents side by side, and everything they write kept as a page you can read, version and search.
The framing on the site is almost poetic for developer tooling: 'Agents make it easy to start. snyvi helps you finish.' Launching at #19 on Product Hunt with 83 upvotes and 63 followers, tagged Productivity, Open Source and GitHub, snyvi is free, open source under the MIT licence, and has no paid tier at all. The product is the whole pitch.
One desk per project
The desk is the whole interface metaphor. Each project gets a desk: its folder, with up to four agents side by side — Claude Code, Codex, Gemini, Cursor, Windsurf, VS Code, Zed, or any shell, since a panel is a real shell that runs whatever you put in it. Each panel tells you whether its agent is working, done, or waiting on you. When one needs a decision from you, it asks in the desk's corner and waits instead of stalling silently in a background tab.
The plumbing is MCP. Any agent that speaks MCP can send documents to snyvi with one tool call, send_document, and setup is one command per agent — snyvi init claude, snyvi init codex, and so on. A local daemon listening on 127.0.0.1:7777 renders each document once and stores it in SQLite under the project it came from. Every open page receives the document over an event stream, about 50 milliseconds after the send. Agents send; snyvi shows.
And the desk keeps its own notes, which is the quiet killer feature. Each desk's notes are readable and updatable by the agents: they say how far they've got and tick one off with the commit it went into. Switch projects and each desk is exactly as you left it — the agents you walked away from keep working. Home shows where every project was left, and ⌘K searches everything any agent ever sent, across all desks.
Nothing scrolls away
The product's second act is memory. Every plan and review an agent writes lands on its desk as a clean page, filed under its project, with every version kept. Nothing scrolls away — the phrase is the site's section header, and it names the exact pain: terminal scrollback is where plans go to die. In snyvi, they're pages: readable, versioned, searchable.
This matters more than it sounds because versioned pages change the trust equation with agents. When an agent rewrites a plan at 2am, you don't wake up to a mystery — you wake up to a diff. The page marks which panel wrote it. The versions accumulate into a project's working memory that survives context switches, machine restarts, and your own forgetfulness.
It's a deliberately small idea executed thoroughly: take the most ephemeral artifact in agent-assisted development — the stuff agents write while you look away — and give it the permanence and structure of a document. The desk metaphor earns its keep here. A folder of Markdown files doesn't tell you what changed; a desk with versioned pages does.
Private, fast, and one honest ask
snyvi's privacy story isn't a policy page; it's architecture. The app binds to 127.0.0.1 and nowhere else. There is no account, no telemetry, no cloud sync. Nothing of yours leaves your machine unless you send a page to a friend you've paired with, sealed to their key. For a tool that watches everything your coding agents do — your half-finished ideas, your client work, your notes — local-only isn't a feature, it's the precondition for trust.
The engineering reflects the same discipline. One static binary, 17.4 MB, written in Rust. Cold start of 11 milliseconds to first health. Updates are signed and checked before they're applied. It ships for macOS, Linux and Windows — with honest rough edges the team discloses upfront: the Windows installer isn't signed yet, so SmartScreen asks once; the Mac build is signed and notarized by CI.
Which brings us to the launch post's most human moment. 'One honest ask: we need a Mac tester,' Sunny wrote. 'The Mac build is signed and notarized by CI, but neither of us owns a Mac. If you do, brew install --cask snymrova/snyvi/snyvi and tell me what the first minute was like. I'll fix what you hit today.' A small indie team shipping to a platform they can't touch, fixing what testers hit the same day — that's the whole ethos in one paragraph. Free, open source, no paid tier: the ask is participation, not payment.
The furniture for the agent era
What's interesting about snyvi is that it's a second-order product: a tool for the era when everyone has coding agents, built by someone who runs a studio shipping real products with them. mRova's own catalogue — a boarding platform, a dog-care app, a supplement brand past 5,000 orders — suggests a team that builds for production, not demos. snyvi v1.46.1, released October 19, 2026, carries that operator's DNA: the installer story is sorted for three platforms, the checksums are verified, the docs explain what is and isn't signed.
The deeper bet is that the agent era's missing layer isn't smarter agents but better desks — working environments that remember. Every coding-agent vendor is racing to make agents more capable; almost nobody is building the furniture around them. snyvi's wager is that the developer with four projects and twelve agent sessions doesn't need a better model. They need to know, at a glance, what's working, what's done, what's waiting, and where the plan went.
Launch day suggests the itch is real. Eighty-three upvotes put it at #19 for the day, and the comment section immediately started doing product discovery — one commenter asked whether snyvi works with agents that don't run in a terminal, like the Claude and ChatGPT desktop apps. Sunny's launch post had anticipated exactly that kind of question: 'how do you keep several agent-built projects straight today? That answer shapes what I build next.' The roadmap, in other words, is being written in the comments. That's how you want a launch to go.
The playbook
- Scratch your own scatter — snyvi exists because Sunny's own work scattered first. The most convincing developer tools are the ones the maker needed on Monday morning.
- One desk, not one window — Not tabs, not windows — desks. A project's folder plus its agents side by side, each panel showing whether its agent is working, done, or waiting on you. The spatial metaphor does the organizing.
- Turn agent output into pages — Any agent that speaks MCP calls one tool, send_document, and the daemon renders it once, stores it in SQLite under the project. The file stays yours; the reading stays easy.
- Local-only by construction — Binds 127.0.0.1 and nowhere else. No account, no telemetry. Privacy as architecture beats privacy as a settings toggle.
- One honest ask per launch — The Mac build was signed and notarized by CI, but the team didn't own a Mac — so launch day's one ask was a Mac tester, not a feature list. Ask the crowd for exactly one thing.