What it does
Task evals per category, like AI evals: good at search, not at highlights. Each result links to a recording.
Install facts: does it start on a clean box, how fast, in how much RAM.
Browse, search, preview and fork self-hostable tools, each on one standard page. See what a tool does, whether it's safe to run, and who stands behind it.
Not open yet. The sample tools below show real, dated facts. We haven't evaluated or scanned any of them yet.
We never award a "safe" badge. We show what each check found, with the logs, and you decide.
How it works
Task evals per category, like AI evals: good at search, not at highlights. Each result links to a recording.
Install facts: does it start on a clean box, how fast, in how much RAM.
Shared checks, run once per release instead of by everyone who installs it:
Rerun on every release, with the diff.
Publisher and account age. Maintainers. A large code dump from one person.
Signed releases. How past security reports were handled.
Every tool gets the same page layout, so two tools compare side by side. Fork any tool from its page.
Sample tools
All eight came up in "Ask HN: What do you run on a $5 VPS that's worth keeping online 24/7?". Each card uses the standard layout.
Fetched 12 Oct 2026 from the GitHub, Codeberg and OpenSSF Scorecard APIs, and from the HN thread (656 comments). Nothing here has been evaluated or scanned by us yet.
Back up and browse your photos and videos, with phone apps.
Stream your own films, shows and music to your devices.
Save web pages as clean, readable text to keep and search.
Lightweight server monitoring with history, Docker stats and alerts.
Check sites and services on a schedule and alert when one goes down.
Git hosting with code review, issues, packages and CI.
Keep folders in sync directly between your own devices.
A self-hosted control server for Tailscale clients.
What an eval looks like
Each category gets a fixed set of real tasks. Every tool runs the same tasks on a clean box, and every cell links to the proof.
Illustration: not real results. Tools A, B and C are made up.
| Task | Tool A | Tool B | Tool C |
|---|---|---|---|
| Save 20 articles, clean text extracted | ✓Pass | ◐Partial | ✓Pass |
| Full-text search | ✓Pass | ✓Pass | ✗Fail |
| Highlights | ✗Fail | ✓Pass | ◐Partial |
| Export your data | ✓Pass | ✗Fail | ✓Pass |
| Share from phone | ◐Partial | ✓Pass | ✓Pass |
| Install on a clean 1 GB box | |||
| Time to first page | 41 s | 3m 12s | 58 s |
| RAM after start | 96 MB | 610 MB | 140 MB |
✓ Pass · ◐ Partial · ✗ Fail · Every cell links to its proof
Read it like an AI eval. Tool A is good at saving, search and export, but has no highlights.
Tool B is good at search, highlights and sharing, but can't export your data. Tool C is light on RAM and has no search.
On a real tool page, each cell opens the recording and log of that run. Here they open this note, because the results are made up.
Verified on launch day
New self-hostable tools show up on Show HN every day. The plan: when one is posted, we check it within the hour and publish the report.
We pick it up from the Show HN feed.
Clean-box install, shared security checks, outbound connections, who published it.
Logs and recordings in the standard layout. Rerun, with a diff, on each release.
This is planned, not running yet. Want the first reports?
Why we're building it
The idea came from an Ask HN thread about what people run on a $5 VPS. Hundreds of replies, and more tools on Show HN every day.
Picking one still means reading READMEs, scanning releases yourself and hoping. We self-host too, so we'll use it first.
Our own app suite will be the first thing through this pipeline. You can try its live demo now.
A stook is a group of sheaves stood upright together in a field: many small things, standing together.
Early access
Leave your email and we'll write when the catalog opens, with the first tool reports.
Privacy. This is all we store: