The Demand Report — Issue #14
Last week's agent record wasn't a spike: this
week came in within four percent of it. On the human
side, people stopped asking for categories and started
asking for their tools by name. And the gap list grew
threefold, led by one industry this catalog barely
covers.
Week of October 12 – 18, 2026 · 31,361 providers ·
155,746 APIs
Every Monday I read what people and agents actually typed
into
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.
📈 The Demand Report
MCP searches came in at 8,629 — down 4% from last
week's record of 8,978, and above every other week
this report has measured.
The series: 4,754 → 5,551 → 3,508 → 8,978 →
8,629.
Every earlier MCP record in this run was followed by a
fall: 4,533 to 1,591, then 5,551 to 3,508. This one
wasn't. Two straight weeks near nine thousand is the
first time agent search has held a new level instead of
spiking. Share held too, at 1.34% of all requests.
The rest of the scoreboard moved the healthy way, which
hasn't happened often. Total requests dipped
673,315 → 644,218, down 4%, and distinct
terms rose 5,110 → 5,237, up 2.5%.
Keyless callers rose 13%, to 554,808, on the same broad
vocabulary.
|
|
This week vs last: total searches −4.3% to 644.2k,
keyless API +13% to 554.8k, unique terms +2.5% to
5,237, MCP searches −3.9% to 8,629. Footnote: agents
held within 4% of last week's record; fewer
requests, more distinct questions
|
People asked for their tools by name
For three weeks the human list was about money, in
categories: payments, refunds, crypto. This week it read
like somebody's software stack:
|
Term
|
Oct 12–18
|
|
Crisp
|
22
|
|
Airtable
|
19
|
|
OpenWeatherMap
|
19
|
|
Eventbrite
|
18
|
|
Typeform
|
18
|
|
SMSAPI
|
17
|
|
Reddit
|
15
|
|
HubSpot
|
13
|
|
Mailchimp
|
13
|
|
Merge
|
13
|
|
Coda
|
12
|
|
Twilio
|
12
|
None of those was in last week's top hundred.
Together they're forms, events, chat, email, SMS, CRM
and docs — the tools a small business wires together. Next
to them:
payments accounting
and
payments billing crm accounting, both at 10. That reads like someone describing a whole
back office in one search.
The money questions didn't go away.
refunds
came in at 54, a third straight week in the top
hundred
(48, 59, 54). Stripe was 25.
The other names at the top:
quietforge
(110, up from 49 — a document-processing API in the index
as a provider with no API pages yet);
Westfund Health Insurance
(38), an Australian health fund;
Hinduja Leyland Finance
(28), an Indian vehicle lender; and
intelipost
(28), Brazilian shipping software. The five-term
resilience stack is still there, all five between 9 and
16.
The gate changed leaders
Bounces off the paid surface rose
60,502 → 64,012.
For the first time since early September, cohorts
isn't the most reached-for gated resource.
security
overtook it, 23,334 → 32,117, while
cohorts
fell to 20,327. And a new resource jumped into third:
resolve, 31 → 4,101. That's the call that
turns a name, domain or URL into the right provider
record. Nobody tried it at volume before this week.
Read the top three together: what does this API's
security look like, which group does it rank in, and which
company is this name. That's what someone doing
diligence asks, not someone browsing.
The authenticated tiers stay off the scoreboard.
pro
reads 18,626, and the tag-quality pass ran on my own
Understanding key on the 17th.
🕳️ The Gaps (demand we couldn't answer)
The pipeline saw 1,671 raw unmet terms,
dropped 971 as noise and 291 the catalog can now answer,
and kept
409 — nearly three times last week's 139.
That's the biggest week of genuine misses in a month,
and the names in it are unusually coherent.
Eight rows, each re-probed against the live search this
morning and still returning nothing:
|
|
The Gaps — top zero-result searches: intelex and
scottish pacific at 5×; enablon, enhesa, ideagen and
wood mackenzie at 4×; bureau veritas and veson
nautical at 3×
|
-
intelex (5×) — environment, health
and safety software.
-
scottish pacific (5×) — Australian
invoice and trade finance for small businesses.
-
enablon (4×) — Wolters Kluwer's
EHS and ESG platform.
-
enhesa (4×) — regulatory compliance
intelligence: which environmental and safety rules
apply where.
-
ideagen (4×) — UK quality, audit and
compliance software.
-
wood mackenzie (4×) — energy and
natural-resources research and data.
-
bureau veritas (3×) — testing,
inspection and certification.
-
veson nautical (3×) — commercial
maritime software for chartering and voyage
management.
Five of the eight are one industry: environment,
health, safety and compliance.
Intelex, Enablon, Enhesa, Ideagen and Bureau Veritas are
the software and the certifiers a regulated company uses
to prove it's following the rules. Not one is in the
catalog. Further down the list, the same lane continues:
codebeamer
(requirements and compliance for regulated engineering),
centrik aviation
(aviation safety management),
aquatic informatics
(water-quality data),
anglian water. Somebody spent the week looking for the compliance
layer and found nothing.
croxyproxy
dropped off the list after four weeks.
If you build in one of these lanes, this is your
invitation.
Add your API in about two minutes:
apis.io/add
— or point your agent at the
apis.io MCP server
and let it submit for you.
🆕 New to the Index
The deployed index reads
31,361 providers and 155,746 APIs —
up 1,373 providers and 8,174 APIs, the
third straight week above a thousand new providers. As
always, the API line counts specification files, not
products.
MCP entries rose 6,591 → 6,966. The hosted-endpoint line
reads 3,319, but it has swung in both directions over the
last month as the index reclassifies how each server is
delivered (2,499, 2,508, 2,278, now 3,319). I'll
treat that line as settled once it holds for two weeks.
Twenty posts went up on the blog last week. All on the
apis.io blog.
⭐ Rated This Week: Airtable — agent-native on points,
held back by one gate
This week's angle: Airtable was one of the tools
people asked for by name, on a week when agents searched
near the record. It's also the cleanest example in
the catalog of a provider that scores into the top agent
band and isn't allowed in.
|
|
Airtable
|
|
Kin Score
|
63.0 strong
|
|
Agent Readiness
|
47.5 agent-ready — gated from
agent-native
|
|
Contracts / callable
|
12 / 100%
|
Agent-native starts at 38.7. Airtable scores 47.5, well
clear of the line, and is held at agent-ready because the
band has two gates. A provider has to document idempotency
and error semantics. Airtable has the second and not the
first.
The door is one of the best I've checked.
mcp.airtable.com/mcp
is Airtable's own hosted MCP server. An anonymous
initialize
gets the right refusal:
HTTP/2 401 www-authenticate: Bearer error="invalid_token", error_description="Missing Authorization header", resource_metadata="https://mcp.airtable.com/.well-known/oauth-protected-resource"
The RFC 9728 document behind it names the resource, the
authorization server, and seven scopes,
from
data.records:read
to
schema.bases:write. A made-up path returns 404. Two weeks ago I wrote that
Square's MCP door advertises no scopes at all.
Airtable's tells an agent, before it asks, exactly
what it could be allowed to do — read records, write
records, change schema, read comments — and lets the human
approve only part of that. Dynamic client registration is
on.
So what's missing is idempotency, and Airtable is
closer than the score says.
Its docs describe an upsert mode for record writes: set
performUpsert
with
fieldsToMergeOn, and a write that matches an existing record updates it
instead of creating a duplicate. Repeat the call and you
get the same record. That's the property an agent
needs for a safe retry. But it lives in prose on a docs
page. Airtable doesn't publish an OpenAPI contract,
so no machine-readable description says so, and an agent
reading a contract can't find it.
The cheapest points on the table are a published
contract that describes performUpsert. That one change would take an
agent-native score out of the gate and give it the band.
The other gaps are the usual ones: no
/.well-known/api-catalog, no delegated identity.
Kin Score rubric 0.23.0, scored 2026-10-11. Bands:
exemplar 66.5+ · strong 54.3–66.4 · developing 39.3–54.2
· thin 26.2–39.2 · emerging 11–26.1 · minimal 0–10.9.
Agent Readiness: agent-native 38.7+ (gated on
idempotency AND error semantics) · agent-ready 28.6–38.6
· agent-aware 5.1–28.5 · human-only 0–5.
🤖 Reproduce this yourself
Airtable's bands, no key required:
find_providers(providers=["airtable"], fields=["slug","score","agent_readiness","api_count"])
Then the door and its scopes, the way I checked them:
curl -s -D - -o /dev/null -X POST https://mcp.airtable.com/mcp \ -H 'Content-Type: application/json' \ -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{}}' \ | grep -i www-authenticate
curl -s https://mcp.airtable.com/.well-known/oauth-protected-resource | jq .scopes_supported
Or read the demand data directly:
apis.io/search-terms/, refreshed every Monday.
31,361 providers. 155,746 APIs. 8,629 agent searches —
the second week at a level that used to be a
spike.
See a gap that's yours? →
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.
You are receiving this because you subscribed to this
newsletter.
Unsubscribe
606 W 57th New York, NY 10019
|