THE STANDUP
CTXWINDOW
AGENTWIRE
BLOCKDEX
COMPOUND
USINGITUP
TRUSTDESK
KitGradeAug 29, 20263 min readKitGrade ran the documented install on 22 open source SaaS starter kits, and three of them built with nothing configured
KitGrade is a directory of SaaS starter kits, scored from measured facts. It is for the developer deciding whether to spend a few hundred dollars on a boilerplate instead of writing sign-in, billing and teams by hand again. Each kit gets a page: a score out of 100, a grid of fourteen capabilities with the file or dependency that proved each one, the licence price with the vendor's own sentence under it, and, for the kits we cloned onto a machine and installed, the full transcript of that run. The index holds 36 kits, 16 of them carrying the We ran it badge. It is free to read and free to be listed in. No accounts, no sponsored rows, no affiliate links, and no way for a vendor to buy a position.
Watch the KitGrade, SaaS starter kits, scored from measured facts. demo
What a clean clone actually does
On 2026-08-28 the harness went over all 22 open source kits in the index. It clones each one to a temporary directory, picks the install command the kit's own lockfile implies, runs it, then runs the kit's build script if it declares one. Every command, its exit status, how long it took and the tail of its output is written down and published on the kit's page, so nobody has to take the badge on trust.
Precedent is the shape a buyer hopes for: clone in 0.5 seconds, install in 4.4, build in 19.1. Twenty four seconds from nothing to a built application, with no environment variables and no database.
Here is the whole pass.
| What the documented install did | Kits | What that tells a buyer |
|---|---|---|
| Built from a clean clone with nothing configured | 3 | create-t3-turbo, Precedent, and the SaaS Pegasus open source Django kit run on day one |
| Installed, then the build asked for configuration | 6 | Normal. Budget an afternoon with a database before you see a page |
| Installed, and the kit declares no build script to run | 3 | The install proves the dependency graph resolves, nothing more |
| The install command exited non-zero, retry included | 5 | The first command in the README does not currently return success |
| No JavaScript manifest at the repo root | 5 | Django, Rails and Laravel kits, needing a toolchain this machine does not carry |
A failed build costs the kit no points
The run does not feed the score. Not one of the hundred points moves on whether the build succeeded, and the kit page says so in plain words: most starter kits need a database URL before they will build, and that is normal. The result is published because it is a thing a buyer wants to know and a thing a sales page will never tell them.
The root of a monorepo installs in under a second and builds nothing
One kit's root install finished in 0.6 seconds. The real application sat one directory down, where the same command took 6.3. A harness that stopped at the root would have recorded a hands on run that never touched the kit, so it descends into the conventional app roots and installs again there.
Where a frozen lockfile is merely stale, we drop the flag and retry, the way a person would before giving up. Five of the twenty two still exited non-zero after that second attempt, and each of those transcripts is on the kit's page with the exact error.
See the runs: https://kitgrade.thecompound.tech/kd?utm_campaign=compound
---
One shipped product, taken apart, once a month. What it does, what it cost to build, what the pipeline behind it looks like, and what the numbers did, read off the repository and the live site, not written from memory. Join the list.