About toolweight
12 comparisons, 156 tools, no money changing hands.
toolweight compares developer infrastructure — sandboxes, storage, deployment platforms, model APIs, registrars, messaging — and lets you weight the comparison by what your project actually needs rather than by what a vendor wants to be measured on.
It exists because the category it sits in is broken in a specific, boring way. Search "best email sending API" and the first page is affiliate content. The ordering is a payout table. The "cons" section for the top-ranked tool is always a compliment in disguise. None of it is illegal and most of it is disclosed in six-point grey text at the bottom, and it still makes the result worthless, because the one thing you needed from the page — an ordering you can trust — is the exact thing that was sold.
Credibility is the moat
A comparison site's only asset is whether a senior engineer believes it. That belief is expensive to build, trivially destroyed, and impossible to buy back. It is also, conveniently, the same asset that makes a site worth citing — by a person writing an architecture decision record, and increasingly by a model answering a question about which sandbox provider supports GPU workloads.
So the strategy is not "be independent for now, monetise later". The strategy is that independence is the product, and anything that trades a little of it for revenue trades away the entire thing. There is no house tool. There is no partner tier. There is no version of this site where a vendor's cheque changes a row's position, because there is no mechanism by which a vendor could write one.
How that shows up in the build
- No affiliate links. Every outbound vendor link is a plain link. There is no tracking parameter, no redirect hop, no commission.
- No sponsored placements, no paid inclusion, no featured rows. Roster membership is an editorial call, published in the open, and argued about in public issues.
- No advertising. Consequently, no reason to block AI crawlers — robots.txt explicitly welcomes them, because being cited is the distribution.
- No accounts, no email gate, no newsletter wall. The changelog is the newsletter and it is an RSS feed, which asks nothing of you.
- No analytics that follow you around. There is nothing to sell, so there is no audience to sell.
- Every value is cited. Source URL, verification date and confidence level on every cell, with the stale ones listed publicly.
What it costs, and who pays
Nothing, and me. toolweight is a static site built from files in a git repository; the hosting is rounding error and the research is time. There is no funding, no company behind it and no revenue model waiting in a drawer. If that changes, it will be stated on this page before it changes anything else, and it will not be advertising or affiliate revenue.
The dataset is published under CC BY 4.0 with a JSON API, an MCP server and markdown twins of every page. Take the numbers, use them in a blog post, feed them to an agent, build something better with them. Attribution is the only condition, and a site that gives its data away is a site with nothing to protect but its accuracy.
Where it is written from
Australia, which is why Sydney region availability, AUD pricing and .au support are first-class comparison fields rather than footnotes. Most infrastructure comparisons are written as though latency and data residency stop mattering somewhere over the Pacific. Australian SMS providers get their own category for the same reason: it is a genuinely awkward market, the vendor documentation is uniformly evasive about it, and somebody should just write the numbers down.
Getting it wrong
This site will contain errors. Prices change the week after they are verified, limits move without announcement, and some facts were inferred from an error message at eleven at night. The honest response is not to pretend otherwise but to make every claim checkable: read the methodology, check the source link on any cell that matters to you, and if it is wrong, say so. Corrections are prioritised above new content, and vendors correcting facts about themselves go through the same public queue as everyone else.