DevSly by the Numbers: 182,641 Tool Runs Across 147 Countries

We have been running DevSly's tool suite in the open for just over a year. Rather than guess at what people find useful, we pulled the actual usage table and looked at it honestly — including the parts that are less flattering. Everything below comes from our own database, covering 26 June 2025 through 23 July 2026.

The headline numbers

That works out to roughly eight runs per visitor across the period — people who find a tool that fits their workflow tend to come back to it rather than use it once.

The fifteen most-used tools

We list runs and unique visitors side by side deliberately. Runs alone reward tools that get hit repeatedly by scripts; visitor counts are the better signal of how many actual humans a tool reached.

Two things stand out. First, the top of the list is dominated by cleanup and inspection work — stripping tags out of HTML, looking up who owns a domain, checking what a JWT actually contains. These are the small, unglamorous steps that sit between one real task and the next.

Second, QR Generator is an outlier in the opposite direction: 4,234 runs from only 781 visitors, about 5.4 runs each. People generating QR codes generate a lot of them in one sitting.

What the categories reveal

Efficiency per tool tells a sharper story than the totals. HTML tools produce 29,132 runs from just five tools — roughly 5,800 runs each. Formatters, with eleven tools, average around 1,400. More tools in a category does not mean more use; it usually means the category has been split more finely than users actually need.

JavaScript tools are the standout: four tools, 16,470 runs, about 4,100 each. That category earns its place.

Where people use it from

147 countries appear in the data. Here are the top twelve — and this is where reading only one column would mislead you.

Ireland sits second on runs and near the bottom on people: 35,242 runs from 90 addresses, nearly 392 runs each. That is not a country of unusually enthusiastic developers — it is cloud infrastructure. Several large providers run major regions there, and automated traffic gets attributed to the datacenter's location, not the operator's. The Netherlands and Czechia show milder versions of the same pattern.

Rank by people instead of runs and the picture changes: the United States (6,476), Singapore (2,313), China (2,060), France (1,106) and Vietnam (1,039) lead. Singapore, China and Vietnam each average under 2.5 runs per visitor — a lot of individuals each doing a small amount of work, which is what organic discovery looks like.

A further 8,015 runs could not be attributed to any country, usually because the address resolved to a private range or the lookup failed.

The five newest tools

Keeping this separate, because comparing a two-week-old tool against one that has been live for a year is meaningless. These are the five most recently added, with everything they have accumulated so far.

cURL Converter has had the strongest start: 602 runs from 245 visitors in about five weeks. Turning a working curl command from API documentation into code for your own stack is a job people clearly do often.

The three watermark removers went live on the same day and are still finding their audience. Worth separating two numbers here: those run counts include everyone who opened the page, while the actual processing jobs are smaller — 25 videos processed through the NotebookLM video tool and 5 through the Gemini one. Video work has a much higher bar than a text utility: you have to have the file to hand, and you have to wait for the encode.

And then there is the XenForo Password Generator: added in October 2025, zero runs since. Not a broken tool — simply one nobody has needed. We are leaving it in the list rather than quietly deleting it, because a stats post that only shows the wins is not a stats post.

The last eight months

Runs have roughly doubled since December while visitor numbers have stayed close to flat, hovering around 3,000 per month. Existing users are leaning on the tools harder rather than a wave of new ones arriving — and part of that gap is automated traffic growing too. July is already the strongest month on record with a week still to go.

The API tells a very different story

Alongside the browser tools, DevSly exposes a small REST API. Its numbers look enormous next to the tools — and they are, but for a reason worth being upfront about.

That last row is the whole story. Essentially all API traffic comes from one integration. 844,132 requests is not 844,132 developers discovering the API — it is one automated consumer, running steadily for a year, plus a handful of others who made a few calls each. We would rather show you that than let a big number imply broad adoption it doesn't have.

Which endpoints get called

The IP-info endpoint alone accounts for 99.2% of every API call ever made. It is a natural fit for automation — cheap, fast at 346 ms, and useful to call on a schedule — which is exactly why one consumer can generate three-quarters of a million requests without anything unusual going on.

WHOIS is the opposite: only 5,419 calls, but at 1,434 ms each it is the slowest endpoint by far, because it depends on external registry servers that we cannot speed up. The webhook management endpoints answer in a millisecond or two — they are just reading our own storage.

Reliability, honestly

Of all API requests, 81.5% were recorded with a 2xx success status. The remaining 18.5% carry no status at all in the log — and that is a gap in our own recording, not 155,785 failed requests. Early in the API's life the logger did not always capture the response code, so we will not dress those up as either successes or errors. It is the one number in this post we cannot stand fully behind, and pretending otherwise would defeat the point of publishing real data.

Traffic was also far from steady. March and April 2026 saw 358,152 and 201,672 requests — more than half the year's total in two months — before settling back to around 30,000 a month. That is the signature of a single automated job changing how often it runs, not of the developer base growing and shrinking.

What we take from the API numbers

The public tools and the API are almost mirror images. The tools show broad, shallow use: thousands of different people each running something a handful of times. The API shows narrow, deep use: one integration leaning on one endpoint, hard, for a year. Both are legitimate — but only one of them is the developer adoption a raw request count might suggest, and it isn't the API. If you are building on the API, there is a lot of headroom.

How we count

So you can judge the numbers for yourself:

• A run is one recorded use of one tool by one address on one day. Opening a tool page counts; the counter increments per day, not per click, so heavy same-day use is undercounted rather than inflated.
• A visitor is a distinct SHA-256 hash of an IP address. We never store raw addresses for these statistics, and there is no cross-tool identity, so the same person on a different network counts twice.
• Country comes from IP geolocation, which reports where the address is registered — for cloud traffic that is the datacenter, not the user.
• Internal counters (rate-limit and queue bookkeeping) are excluded from every figure above.
• API requests are logged separately from tool runs, one row per call, and the two are never mixed in the same total.

What we take from this

The tools that win are boring in the best sense: they remove one specific piece of friction and get out of the way. Nobody plans their day around an HTML-to-text converter, but 2,352 different people reached for one.

The category numbers argue for consolidation over expansion — eleven formatters do not outperform five HTML tools. And the country data is a reminder that a single popular metric can point the wrong way entirely; without the visitor column, we would have concluded that Ireland was our second-biggest market.

We will publish this again as the numbers move. If a tool you rely on is missing from the list, it may simply be one nobody else has found yet.