A very fantastic little desktop app
Looper quietly watches the folders you already use, reads every spec and scratchy note, and wires them into one read-only index your whole team — and their agents — can reach. You write. It wires. Finished.
The problem
Have you noticed the lifecycle gap between specs — the stuff used to build — and docs — the stuff that describes what is? Agents get confused about what's being built versus the current state. Sometimes you need specs that govern changes across multiple repos, which means they need a new home. The simple fix: keep only current-state docs with the code, in the same repo. Put specs somewhere else, in their own repo. helps you pull off that pattern with set-it-and-forget sync strategies that keep your specs repo up to date at all times, without managing it like code. In other words, your specs repo can be treated as a glorified fileshare.
markdown files in your folders. You remember writing maybe thirty of them.
Every repo has one. Nobody opens it. The knowledge just quietly rots.
Your search finds the filename you half-remember — never the idea you actually need.
How it works
Pick the folders and git repos you already use. Looper leaves every file exactly where it is — nothing moves, nothing gets renamed.
It harvests every doc into a read-only index, follows the links between them, and keeps it live in the background while you get on with your day.
You, your team, and your agents all read from one source of truth. Search instantly, explore the graph, and watch it stay current as you write.
Keep a git repo of nothing but authoritative specs. Looper auto-syncs it and weights its docs as a higher grade of truth — the canon you and your agents reach for first.
Let an AI provider sharpen the index in place — summarize, cross-link, and raise the quality of the docs your agents read. Off by default; entirely your call.
What's in the box
New doc saved? It lands in the read-only index before you've switched windows. Zero filing, zero ceremony.
A live file watcher keeps the index honest. Close the laptop mid-thought; it quietly catches up on launch.
The Viz view turns your pile of docs into a graph you can actually explore — clusters, links, the works.
Full-text search across everything — ranked, fast, and local. The idea you needed, not just the filename.
Finds the repositories living in your folders and links them with a tidy .looper marker. No mess.
Recent, most-read, and freshly-indexed docs — your whole brain at a glance, the moment you open the app.
The home base
Open Looper and land somewhere useful: what you touched last, what you read most, and what's indexing right now.
The big picture
The Viz view draws every doc and the links between them. Spot the clusters, find the orphans, follow a thread you forgot you'd started.
Bring your own brain
Looper's enrichment features run on your agent — starting with Google's Gemini. Paste an API key once, or just reuse the GEMINI_API_KEY already sitting in the shell you launched from. Your key, your files, your machine.
No accounts. No telemetry. You only ever pay your provider for your own usage. We'll add support for more AI providers soon.
Reuses the GEMINI_API_KEY from the shell Looper was launched from.
Reasonable questions
Yes. Free forever — you only ever pay your AI provider for your own usage, and that's between you and your API key.
The app is Looper: as in, close the loop between AI and humans. The company is Very Fantastic: an upbeat name to bring a little levity to the world.
No. Looper is a local desktop app. Your docs and your API key never leave your machine — there's no server for them to leave to.
macOS, Windows, and Linux. One app, three operating systems — it's built on Tauri, so it stays small and native-ish.
Never. Looper reads in place. The only things it ever optionally touches are .gitignore, .looperignore, and symlinks.
We've started with Gemini because of the value OKF brought to the project. Support for the other big AI providers is coming.
macOS · Windows · Linux · ~12 MB · no account