---
type: "article"
title: "The Demand Report — Issue #03"
newsletter: "APIs.io"
newsletter_handle: "apisio"
newsletter_url: "https://usecommune.com/n/apisio"
author: "Kin Lane (@kinlane)"
published: "2026-08-05T12:00:39.472Z"
canonical_url: "https://usecommune.com/n/apisio/a/cPzbqKAG"
markdown_url: "https://usecommune.com/n/apisio/a/cPzbqKAG.md"
chat_url: "https://usecommune.com/n/apisio/a/cPzbqKAG/chat"
body_source: "native"
likes: 0
replies: 0
body_words: 2189
---

# The Demand Report — Issue #03

*Traffic went up 5×. Demand didn't move. And the ruler I measure providers with changed underneath everyone.*

**Week of July 27 – August 2, 2026 · 25,933 providers · 108,265 APIs indexed**

***

Every Monday I read what people and agents actually typed into [apis.io](https://apis.io/) — the search box and the MCP server — as a demand signal for the API economy. Nobody else has this data, so nobody else can send you this email.

This week the data tried to lie to me twice. Both times the honest number is smaller and more interesting than the headline. Here they are anyway.

***

## 📈 The Demand Report

**123,498 searches, up 225% week over week.** Ignore that number. I'm printing it because I'd rather show you the one I threw out than quietly not mention it.

Requests to the apis.io API are logged by tier. `internal` means no API key and an apis.io origin — the site calling itself, every page render, every crawler walking those pages. `keyless` means somebody hitting the API directly from outside. The split:

|   | Last week | This week |   |
| --- | --- | --- | --- |
| Internal (the site itself) | 32,611 | **118,232** | \+263% |
| Keyless API (outside callers) | 5,404 | **5,266** | −2.6% |
| MCP server | 1,027 | **518** | −50% |

![This week vs last: total searches +225% to 123.5k, keyless API flat at 5,266, unique terms −8% to 5,595, MCP searches −50% to 518](https://fmxorstnlyoqpqdftivs.supabase.co/storage/v1/object/public/article-images/15d2e398-36ac-48cf-81d9-7b3858493d21/0e295ca7-0abe-4fdc-bea5-344b560d5529/1785931120015-qam0gj0r.png)

**All of the growth is the site being crawled.** Google Analytics puts the week at 502,881 sessions against 95,693 the week before. **422,260 of them — 84% — came from Singapore**, up from 37,859. Vietnam contributed another 44,185. Average session: 12 seconds. Sessions peaked at 170,183 on July 29 and collapsed to 37,634 the next day. CloudFront requests rose 50% to 9.5 million and the error rate went from 10.55% to **16.49%**, which is what a bot walking URLs that don't exist looks like from the edge.

United States sessions over the same week: 33,189 → 37,873. Up 14%. That's the human line, and it's the only one I'd put weight on.

**So here is the actual week.** Outside demand was flat. Unique search terms fell 8%, from 6,082 to 5,595. And the line I led with last issue went the other way.

**Last week I told you agent searches tripled — MCP traffic up 231% to 1,027. This week it halved, to 518.** One week is not a trend, and neither was the up-week. I'd rather hand you the reversal than only report the direction that flattered the thesis. Two data points and a reversal is what I have; when I know why, I'll tell you.

**One more number I'm throwing out.** The top search term this week was *"Automation Preflight API"* — 1,185 times, against 96 for the runner-up. It resolves to exactly one fuzzy match in the whole catalog (Adobe InDesign, on the phrase "preflight automation" buried in a description). A single term at twelve times the volume of everything else, pointed at one weak match, is a machine in a loop, not a person shopping.

**What's left, after all that, is a boring and genuine top-of-list**: SCIM (27), Comcast (18), Stripe (18), Twilio (15), weather (16), football (16), payments (10), Chart of Accounts (10), Digital Asset Management (9), Permission Sets (10), Shipping Accounts (9), Allow Lists (8). Identity plumbing, the usual integration primitives, and a long tail of people with a specific enterprise object to move. That list barely changes week to week, and that stability is the point: the agentic layer is being added on top of this, not instead of it.

**And the Pro-tier signal, which is its own kind of demand:** 29 attempts on `ratings`, 15 on `changes`, 9 on `me`, 5 on `security`, 4 each on `scopes` and `compare`. People keep reaching for the scoring data and the change feed specifically.

***

## 🕳️ The Gaps (demand we couldn't answer)

279 distinct terms came back empty, up from 263. Supply grew — 494 providers and 1,339 APIs added this week — and the gap grew anyway.

*A note on method: the raw unanswered list had 695 terms; 416 were dropped as noise. That filter is imperfect and I can see exactly how: it strips single common English dictionary words using the system US wordlist, so this week's crawler got through on British spellings — Anaesthetise, Civilisation, Disincentivise, Fantasise, Jeopardise, Agonise. Fixable, and now on the list.*

**The two loudest unanswered searches this week weren't categories. They were domains.** `archlending.com` (21×) and `chainfi-finance` (21×) — each searched roughly two dozen times, each returning nothing every time. That is the shape of a company typing its own name into an index to see whether it's there, and finding out it isn't. If that was you: it takes about two minutes to fix, and the link is at the bottom of this section.

![The Gaps — top zero-result searches excluding two domain self-lookups: Pismo 8×, Insurtech, SoftPro Sync and Zoomrooms 4× each, Airfare 3×, Bedrock AgentCore, ShipBob and Autonomous Vehicles 2× each](https://fmxorstnlyoqpqdftivs.supabase.co/storage/v1/object/public/article-images/15d2e398-36ac-48cf-81d9-7b3858493d21/0e295ca7-0abe-4fdc-bea5-344b560d5529/1785931155983-42ggghys.png)

**The sharpest real gap is Pismo** — searched 21 times, empty 8 of them. Visa-owned core banking and card processing, and the index has nothing. That's a genuine hole in a sector I write about constantly.

**Named products nobody has submitted:** SoftPro Sync, Zoomrooms, Bedrock AgentCore, ShipBob, anam.io, coinglass, statseeker, inpay, logmate, airnow, Adyen Hosted. Someone specifically wanted your API this week and got nothing.

**Sector slang the catalog doesn't speak:** Insurtech (4×), Healthtech (2×), legal tech (2×) — all empty. People search by the word the industry actually uses; the catalog indexes by something more literal. That's a vocabulary gap I own, not one a provider can fill for me.

**Categories with no coverage:** Airfare, Autonomous Vehicles, Fusion Energy, Circular Economy, Civic Technology, Credit Union Offices, Tribal Governance, Fluoropolymers and Fluorochemicals.

**And one I have to own outright.** *"Greater China"* was searched twice and returned nothing — on a week when apis.io published [an analysis of API coverage across Greater China](https://apis.io/2026/08/02/api-coverage-across-greater-china/). The post exists. The search box can't find it. That's a defect on my side, not a gap in the market.

**A methodology finding worth naming**, because it changes how you should read any zero-result list including mine: the unanswered log captures search-as-you-type. `apifre → apifrea → apifreak → apifreaks → apifrek → apifreka` is six entries at 2× each. It's one person, typing, once. Same with `Pygeo → Pygeoa → Pygeoap` and `commoditypric → commodityprice`. Nobody has deduplicated keystroke ladders out of a public demand feed, so treat the low-count tail everywhere — mine included — as noisier than it looks.

**If you build in one of these lanes, this is your invitation.** Add your API in about two minutes: [**apis.io/add**](https://apis.io/add) — or point your agent at the [apis.io MCP server](https://apis.io/) and let it submit for you.

***

## 🆕 New to the Index

The catalog grew by **494 providers and 1,339 APIs** this week, to 25,933 and 108,265. Two provider profiles published since the last issue:

- **Merge** — 118 APIs, 117 specs; Kin Score **75.3 (exemplar)**, Agent Readiness 63.5 (agent-native). Ten collections, and each one is a *category* rather than an endpoint group — accounting, ATS, CRM, HRIS, ticketing, file storage, chat, knowledge base, plus a dedicated Agent Handler surface *(Unified API)*
- **Kong** — 139 APIs, 132 specs; Kin Score **63.4 (strong)**, Agent Readiness 59.6 (agent-ready). Two collections across 139 APIs, the coarsest decomposition in the sweep *(API Infrastructure)*

Kong is the one worth sitting with. **The company whose business is helping others run their APIs well scores strong, not exemplar** — below Adyen (77.8), Palo Alto Networks (75.3), Merge (75.3), FactSet (73.5) and Slack (73.1) from the same sweep. The reason is architectural rather than negligent, and it's in the full profile.

Also published this week: the MCP server index, the Y Combinator portfolio's API footprint, OAuth scopes across the catalog, and the open-source tag. All on the [apis.io blog](https://apis.io/blog).

***

## ⭐ Rated This Week: I changed the ruler

No provider this week. The rubric itself is the story, and if I ranked somebody on top of it without saying so first, the ranking would be worthless.

**Between last issue and this one, the Kin Score went from 0.5.1 to 0.9 — four releases, all on July 31 — and the whole index was re-scored on August 3.** Nobody's API changed. The measurement did. Here's what moved and why, because a score you can't audit is just an opinion with a number attached.

**0.7 — security defects a contract declares about itself.** Three new checks, measured first across 14,195 as-published OpenAPI documents: contracts declaring the OAuth implicit grant (775 of them) or resource-owner-password (72), both removed in OAuth 2.1; API keys passed `in: query` (815), where the credential lands in logs and referrer headers; and OAuth schemes that enumerate no scopes at all — **only 15.3% of contracts enumerate any**. All three score the *presence of a defect*, not the absence of a feature, so a bearer-token API passes all three vacuously. 25,574 providers re-scored, 0 errors.

**0.8 — provenance grading, finally applied to everything.** We write OpenAPI specs for providers who haven't published their own. Until 0.8, only one check discounted that; the other 112 points of contract quality credited providers for *our* craftsmanship. Now the whole block is graded down to a 0.25 floor when the corpus is ours. Measured against the live catalog before shipping: **318 providers moved. All 318 fell. None rose.** 187 changed band, including **16 that dropped from exemplar to strong**. The canonical case: Yardi, 57.7 → 49.6, strong → developing, because its entire spec corpus was ours and it had been strong on the strength of it.

**0.9 — the dimension was named after an artifact and measured the wrong thing.** `asyncapi_events` only looked in the `asyncapi/` directory. But OpenAPI 3.1 has a top-level `webhooks` object and `callbacks` has existed since 3.0. Across the verbatim harvest archive, **125 providers describe their event surface that way** — 87 via `webhooks` covering 2,218 individual events, 39 via `callbacks` covering 174 — against **zero** who publish AsyncAPI. So 100 providers were scoring `false` on a dimension they satisfy, in the contract they actually maintain. Renamed `event_surface_described` and widened.

**What that did to last week's headline.** In Issue #02 I told you Alpaca was the top-rated provider in the index at 86.4. Today the catalog reports **85.2**, and it is no longer first. Two of its facets went *up* over that window — governance 60.5 → 68.8, contract quality 69 → 71.3 — and the composite still fell, because discoverability (100 → 92.6) and the conditional securities-regulatory layer (100 → 86.7) are measured against denominators that got longer. Alpaca did nothing wrong. It's a very good API profile. It's being graded harder.

**And it will move again.** Kin Score 0.9.1 shipped this morning: the scorer was classifying specifications by looking for an `openapi:` declaration, so a Swagger 2.0 document — which declares `swagger: "2.0"` and uses `definitions` and `securityDefinitions` — was dropped before a single check ran. **A provider whose entire corpus is Swagger 2.0 was scored as publishing no contract at all.** 190 such documents across 90 providers in a 3,024-file sample, Microsoft Azure heaviest, then Kaltura, Mailchimp, Mastercard and Zuora. Rebuilding the index with the fix took it from 81,867 to 87,680 documents across 6,997 to 7,200 providers, and cut parse errors from 800 to 126. Those corrected scores are held until the next full rebuild, so what you see on apis.io today is still 0.9.

It surfaced because Oracle publishes a first-party contract for all 161 of its Cloud Infrastructure services, 7,918 operations. Harvesting them to replace six specs *we* had written **lowered** Oracle's contract quality — the six models were the only documents the reader could parse. A provider was being scored on our modelling and penalised for its own engineering.

**That's the honest state of it.** The rubric is public and versioned precisely so this is auditable rather than mysterious: [github.com/api-evangelist/kin-score](https://github.com/api-evangelist/kin-score). If your score moved and you want to know which check did it, ask me and I'll send you the line item.

*The bands, for reference: exemplar 70+ · strong 60–69 · developing 45–59 · thin 30–44 · minimal 0–29.*

***

## 🤖 Reproduce this yourself

Everything above came out of public apis.io surfaces. Paste this into your agent (apis.io MCP) to pull the current top of the index — scored on rubric 0.9, as of August 3:

```javascript
find_providers(sort="composite", limit=25, fields=["slug","name","score"])
```

Compare it to the same call next week and you'll see 0.9.1 land.

Or check the demand data directly: [**apis.io/search-terms/**](https://apis.io/search-terms/) — the top 100 searches and the top 100 unanswered ones, refreshed every Monday.

***

**25,933 providers. 108,265 APIs. 5,266 keyless searches that were actually yours. 279 still unanswered.** See a gap that's yours? → [**apis.io/add**](https://apis.io/add)

*The Demand Report is a weekly read of apis.io's own search + discovery signal. Forward it to someone who ships an API.*

***

## Discussion

No replies yet.