What are you trying to find out?

What 1,005 real public pipelines reveal about how teams actually govern their APIs — and the blueprint almost none of them have reached.

What fifty Consumer Data Right banks actually publish, scored — and what it means for the operators who run them and the investors who back them.

What the market data providers actually publish, scored — and where the durable value hides in a commoditizing feed.

What sixty UK banks and building societies actually publish, scored — from the CMA9 Open Banking mandate to the challengers, building societies, and private banks arou...

The twenty-five most important venture funds, ranked on one axis — whether their portfolios are ready for agents that act.

What 113 US banks, credit unions, neobanks, and banking-rails providers actually publish, scored — the one open-finance market with no mandate, and the widest split in...

What 42 Canadian banks, credit unions, caisses, and fintechs actually publish, scored — the emptiest banking market measured, where the rails outrank the banks.

What sixty-eight US payment companies actually publish, scored — the deepest, most agent-instrumented payments estate on earth, and the market with no mandate and no f...

What thirty-two UK payment companies actually publish, scored — the open-banking payment-initiation heartland, and the market that hides its own crown jewels.

What twenty Australian payment companies actually publish, scored — the NPP-native market where the signature resource is PayID, and the agent surface is almost entire...

What fourteen Canadian payment companies actually publish, scored — the thinnest, most concentrated market, where the fintechs out-API the incumbents and the rails.

Thirty years of mandated machine-to-machine standards produced clearinghouses, not APIs. The most standardised market measured here is the least programmable one.

What thirteen UK healthcare organizations actually publish, scored — the single-payer market where the incumbent leads and the challengers orbit.

What twelve Australian healthcare organizations actually publish, scored — where the challengers out-execute the national agency, and one built the incumbents' FHIR.

What ten Canadian healthcare organizations actually publish, scored — the market with no FHIR-native challenger class, where a practice-management vendor tops the field.

What seventy-nine US insurance organizations actually publish, scored — the market with no mandate, where one company in the country publishes bind.

What thirty-five UK insurance organizations and the London Market actually publish, scored — where a 337-year-old market beat every regulator to open APIs.

What twenty Australian insurance organizations actually publish, scored — the country with the machinery for open insurance and no obligation to use it.

What twenty Canadian insurance organizations actually publish, scored — the only market with a real API certification programme, and its largest carrier is Not Yet Rated.

What eighty-three telecom organizations worldwide actually publish, scored — the industry that built a real open standard and then published 351 specifications with no...

What forty-five US real estate organizations actually publish, scored — the only industry that successfully mandated a machine-readable contract on itself, and the two...

What twenty-four UK real estate organizations actually publish, scored — the market with no MLS, no standard and no mandate, and the best-published real estate sector ...

What eighteen Australian real estate organizations actually publish, scored — the market that mandated the rail instead of the schema, and got an API that carries prod...

What fourteen Canadian real estate organizations actually publish, scored — four facets at a clean sweep of zeros, and the fifth consecutive sector in which Canada fin...

What one hundred and six headless CMS, commerce and browser providers actually publish, scored — the architecture whose entire promise is that the API is the product, ...

What twenty-four Australian energy organizations actually publish, scored — proof that a statutory data mandate transplants from one industry to another, and what it c...

What sixty US energy organizations actually publish, scored — the market with the best standard, the largest scale, no obligation, and the lowest average of the four.

What twenty-six UK energy organizations actually publish, scored — the market that mandated the pipes instead of the data, and got exactly what it compelled.

What eighteen Canadian energy organizations actually publish, scored — one province mandated Green Button, and the country's best API belongs to a utility nobody compe...

What thirty US travel and aviation organizations actually publish, scored — the sector where switching cost IS the commercial structure, and the published surface is i...

What fourteen UK travel and aviation organizations actually publish, scored — the country that proved mandates work, in the one sector it never mandated.

What eleven Australian travel and aviation organizations actually publish, scored — the control group, and the lowest contract-quality reading in this entire research ...

What nine Canadian travel and aviation organizations actually publish, scored — the country that authored the travel industry's flagship standard, and the market where...

The AI industry adopted the protocol it invented and none of the safety rails that make it drivable. It shipped the door and skipped the locks.

The best-scoring industry in the catalog — first of twenty-five — and it publishes idempotency at the same 2.8% rate as the worst. The gap was never competence.

The second-best-scoring market in the catalog — and 72% of it will not tell you how to report a vulnerability, now that the EU Cyber Resilience Act requires exactly that.

What thirty-six API management platforms, gateways and data planes actually publish, scored — the market that sells other people their API practice, measured on its ow...

The industry that moves the physical economy publishes idempotency at 1.3% — fifteen companies out of 1,154 — in the one market where a retried request dispatches a se...

The companies that rent the chips score 36.5. The companies that design and fabricate those same chips score 10.7. NVIDIA scores 24.0 and NVIDIA NIM scores 78.7.

The market with the best standard and the most mandates publishes consent as a machine-readable surface at 3.4% — the same rate as the rest of the catalog, in the one ...

One of the best-scoring markets in the catalog — and the segment selling "know your data" is the worst-scoring segment inside it, publishing a contract 16.5% of the ti...

The industry whose tracking wrote the world's consent laws describes consent in its own APIs at 4.4% — below the whole-catalog rate, and at zero across the consent-man...

One of the highest-publishing large markets in the catalog — median 26.4 against 15.0, idempotency at nearly three times the catalog rate — and OAuth scopes published ...

What twenty-nine documentation platforms, reference renderers and docs-as-code frameworks actually publish, scored — the market whose entire job is making APIs legible...

The library out-publishes the classroom — scholarly infrastructure publishes a contract 51.0% of the time, corporate learning 0% — and thirty-four organizations in 810...

The contract lifecycle management segment publishes the fewest contracts in its own market — 18.2% against e-signature's 55.0% — and the market that sells governance s...

What thirty-seven API testing platforms, mocking tools, load generators, contract frameworks and scanners actually publish, scored — the first market in this series th...

The market that consumes APIs rather than publishing them — why only three of thirty-five publish an API of their own, and why that is a deliberate market movement rat...

This market is not short of APIs — it is short of public ones. 5.8% publish a contract, and the most likely reason is a security posture that treats a documented inter...

What forty identity, authentication and authorization providers actually publish, scored — the strongest-publishing market in this series, and the one where the scarce...

The weakest area in revenue software, and it holds at full coverage. Governance averages 18.1 in the market that holds recordings of every customer conversation.

This area publishes contracts at a higher rate than the market around it and still fails every delegation dimension. The data moves; the permission to act on it does not.

Sixty-seven percent publish a contract — the highest of any large area in this report. Ninety percent cannot tell an agent whether a call was already made.

This is the most connected software in go-to-market and it sits 1.2 points below the market median on Agent Readiness. Being integrated by everyone is not the same as ...

The best-scoring area in revenue software on every artifact measure, and it fails consent and idempotency at 91% each. Even the market's integration specialists have n...

Twenty-three companies, scoring above the market on every artifact measure and failing consent at 96%. The step immediately before money moves is the step least prepar...

Twenty companies, and four of the market's own core terms carry no company at all. This is the one area in the report where coverage is a live question.

This market has made itself readable by machines and has not made itself operable by them. Fifty-eight percent publish a contract and forty-six percent run their own M...

One of marketing's strongest areas, with an eleven-point buy/sell split inside a single function. The same capability, measured differently depending on who pays for it.

The lowest governance figure in marketing, in the area whose entire product is an asset other systems depend on retrieving.

The second-strongest area in marketing, and the one where the product only works if the handoff works. Programmability here is not a virtue, it is the business model.

The one area in this report where a low score is partly not the company's fault — and one of only two where the sell side outscores the buy side.

The strongest divide in marketing software is not category, it is customer. Software sold to the advertiser scores 53.3; software sold to the publisher scores 39.0. Bo...

The area that automates marketing scores at the marketing average. Orchestration software that cannot itself be orchestrated is the clearest irony in this research.

The strongest area in marketing, built by companies whose product was always an interface. Even here, delegation is missing at over eighty percent.

This market transacts entirely by machine and describes itself to the outside world less than almost any other. The bid stream is a specification. The platform is not.

The lowest scores in either cohort, and the clearest single finding in this research: the side of the transaction with no customer demanding transparency did not build...

The tightest Kin-to-Agent-Readiness gap in either cohort. What these companies publish, an agent can mostly use — there is simply not very much of it.

The area that connects the two cohorts in this research scores like the stronger one. Software defined by a handoff gets good at handing off.

The newest money in marketing arrives on the least described surfaces. Retail media grew into a major channel without building a public interface layer.

The most consistently contract-publishing market measured anywhere in this research is also one of the least operable. Government did the part the policy named and sto...

This market solved interoperability where devices meet networks and left it unsolved where customers and agents meet platforms. The best contracts in the research seri...

This market proved it will adopt an API standard when one exists at the right layer. It has done that exactly once, for the narrowest problem in the stack.

98% of the upper band publishes a contract — the highest rate measured anywhere — and it is mostly weather APIs. The software that climate regulation created is sevent...

The product is deliberately closed and the supply chain is wide open. Heroic Labs scores 78.8; Take-Two scores 7.1.

The market everyone describes as dependent on platforms turns out to be the one best equipped to survive them.

The best-governed market in this research scores badly on an instrument built for agents that write. That is a finding about the instrument as much as the market.

Forty years of machine-to-machine standards produced spacecraft that talk to each other across agencies and almost nothing a customer's agent can operate.

An industry that built three workflow languages and published zero workflow contracts. Biotechnology solved sequencing, orchestration and reproducibility for compute p...

The best-documented market in this research is the least agent-ready one. Mobility built excellent interfaces for a world of scheduled integrations and webhooks, and t...

The industry that standardised itself without being told to — 1,369 participants, 487 organizations, 93 repositories, nothing graduated, and exactly one operator on ea...

14,257 real OpenAPI documents, read as published. One in three carries no callable address, and OAS 3.2.0 reached four of them in ten months.

Across 26,722 providers, exactly one publishes an AsyncAPI document — and 159 describe their events in OpenAPI instead. In its own strongest segment, AsyncAPI is losin...

9,439,146 schema objects measured inside published API contracts — and an industry still writing with the vocabulary it had in 2013.

812 Arazzo documents exist across all of public GitHub. API Evangelist holds 5,276. And the market adopted the specification as a test format, not a description format.

3,060 providers carry an MCP artifact. 672 publish a server you can actually reach. And the median server exposes 11 tools while the largest exposes 588.

The identity layer under everything, declared in 102 of 14,195 published API contracts — 0.72%.

122 stars, 21 contributors, two people writing four fifths of it — and Elastic, MongoDB, Cloudinary and Vercel running it in production. Attention and adoption have co...

793 providers publish a GraphQL schema defining 69,521 types — the second-largest description corpus measured here, and the one JSON Schema cannot reach.

775 published contracts still declare the implicit flow — 28.9% of every OAuth-using contract in the corpus, for a flow removed from OAuth 2.1.

Eleven of the largest agent builders specified the discovery layer. Nine companies publish it, seven of them are outsiders, none is fully conformant — and two fail on ...

65 organizations publish an Agent Card. Ten publish one a conformant client can read. And the ten that got it right score lower on API maturity than the forty-one that...

Every Andreessen Horowitz company that ships an API, scored the way a machine reads it — and what the book says about the fund's real bet.

Every Y Combinator company that ships an API, scored the way a machine reads it — and why the batch is agent-native and governance-empty at the same time.

The largest growth-equity API book after a16z, scored the way a machine reads it — and why the estate that discovers best is the one least ready for agents.

Every Accel company that ships an API, scored the way a machine reads it — the infrastructure fund whose contracts are the broadest in venture and whose agent-executio...

Every 500 Global company that ships an API, scored the way a machine reads it — a global seed book that is agent-native at the edges and governance-empty at the core, ...

Every Techstars company that ships an API, scored the way a machine reads it — a broad accelerator book of clean early-2010s contracts with a thin exemplar ceiling and...

Every Lightspeed company that ships an API, scored the way a machine reads it — and why the top of the book and its real agent leaders are two different lists.

Every Bessemer company that ships an API, scored the way a machine reads it — the highest mean in venture, and the softest peak behind it.

Every Battery Ventures company that ships an API, scored the way a machine reads it — a deep-contract infrastructure book whose OpenAPI surface outruns its governance.

Every Google Ventures company that ships an API, scored the way a machine reads it — and why a book of category-defining logos still reads agent-shallow behind Stripe.

Every General Catalyst company that ships an API, scored the way a machine reads it — and why a book that shipped agent front doors on banking and health APIs never sh...

Every SoftBank Vision Fund company that ships an API, scored the way a machine reads it — and why the biggest checks in venture buy the thinnest ceiling on the board.

Every Index Ventures company that ships an API, scored the way a machine reads it — a small, dense, enterprise-SaaS book with real agent depth at the top and a diligen...

Every Norwest Venture Partners company that ships an API, scored the way a machine reads it — a growth-stage barbell of Fortune-100 incumbents and a thin agent-native ...

Every Bain Capital Ventures company that ships an API, scored the way a machine reads it — a broad, enterprise-mature book whose one exemplar already exited, and whose...

A small book with the densest top end measured in this series, where every exemplar is a frontier AI or design-infra bet and the mature franchises ship breadth but no ...

Every Redpoint Ventures company that ships an API, scored the way a machine reads it — and why a book that scores near the top of the mean ships almost no agents at all.

Every Menlo Ventures company that ships an API, scored the way a machine reads it — and why a fund that backed the author of the agentic protocol has barely shipped it...

Europe's most agent-native seed book, scored the way a machine reads it — the highest agent-native rate of any fund measured, built on authored skills over shipped MCP...

Every Speedinvest company that ships an API, scored the way a machine reads it — a European seed book that learned to publish contracts before it learned to govern or ...

Every Canaan Partners company that ships an API, scored the way a machine reads it — and why a broad, long-vintage generalist book still has no reference-grade API.

Every Partech company that ships an API, scored the way a machine reads it — a European operator's book that went agent-native before it built a single exemplar.

A small, focused seed book of the infrastructure agents will consume, scored the way a machine reads it — that has not yet been packaged for agents itself.

Every DCVC company that ships an API, scored the way a machine reads it — a small deep-tech book that instruments its own agents but builds almost nothing at the top.

Every Qiming company that ships an API, scored the way a machine reads it — and why a competent China deep-tech book publishes APIs but has not yet operationalized trust.

A curated winners' slice of Sequoia Capital's book, scored the way a machine reads it — the highest quality and exemplar rate in the series, and the one that trails it...

What JPMorgan Chase is actually buying, read from 7,799 of its own job postings — and what its API surface says back.

What Salesforce is actually buying, read from 1,434 of its own job postings — and the exemplar API surface it will judge yours against.

What Autodesk is actually buying, read from 632 of its own job postings — and the exemplar API surface behind it.

The Fortune 1000's single heaviest AI buyer, read from 1,187 of its own job postings — and the API surface behind it.

What ServiceNow is actually buying, read from 343 of its own job postings — and why a platform this API-heavy has no downloadable contract for you to design against.

What Palo Alto Networks is actually buying, read from 1,207 of its own job postings — and why the product line it hires hardest around is the one it publishes no contr...

What Oracle is actually buying, read from its own job postings — and why one of the largest published API surfaces in the Fortune 1000 was effectively invisible to mac...

What Avalara is actually buying, read from all 108 of its own job postings — and why a company that publishes 2,469 markdown documents for AI agents has machine-readab...

What Postman is actually buying, read from all 116 of its own job postings — and why the company whose product is API specification tooling publishes a specification o...

What Workday is actually buying, read from all 341 of its own job postings — and why the only machine-readable contract you can download from it without a login is SOAP.