Copy page

Account

Plans & Pricing

The APIs.io API and MCP server are free to read. The four plans are named for what you are doing with the catalog, and they follow one another in order:

• Explore — where you begin. Search the catalog and read any single record in it: find a company, read its profile, its APIs, its artifacts, its onboarding, the capabilities and agent-readiness dimensions it satisfies, the provenance behind its score, and the headline Kin Score and agent-readiness score its public page already shows. Browse tags, the taxonomy, and every artifact collection. No key or plan required.
• Learn — where you widen that out. The same surface as Explore, with your own identity and more room to work in. No cost, and one GitHub sign-in away.
• Understanding — where you develop a deep read of the catalog: anything that synthesises across records rather than reporting one. Ranked lists and leaderboards, the scoring rubric, cohorts, the capability model, demand insights, stack design, the venture layer, movement over time, and the Saved Workspace.
• Influence — where you shape your own entry in the catalog. What you fail and what it is worth, which bands are closed to you, what a set of fixes would actually score, and asking a person to re-profile you or correct what we hold. Research about other people's APIs is not here; it is all in Understanding.

Two rules follow from that, and they explain most of the table below.

Cardinality. Everything about one provider is free, including its scores. The numbers across a set are the product: bands stay free everywhere, so you can still find every exemplar payments provider, filter by score and sort by quality — but a free plan returns bands where a paid one returns numbers, and cannot walk away with the scored dataset.

Subject, not cost. The line is what a question is about, not how expensive it is to answer. Ranking the catalog by score is Understanding; choosing between two named vendors is Influence.

	Explore	Learn	Access	Understanding	Influence
Price	$0	$0	per call, from $0.005	$199/mo	$499/mo
Auth	none, or an agent principal's key	GitHub login	GitHub login + card	GitHub login	GitHub login
Rate limit	5 req/s	20 req/s	100 req/s	100 req/s	400 req/s
Daily quota	500	2,000	10,000	10,000	25,000
Search · providers · APIs · tags · artifacts	✅	✅	✅	✅	✅
Playground — curated no-key APIs to learn on	✅	✅	✅	✅	✅
MCP server	✅	✅	✅	✅	✅
One provider's onboarding · capabilities · evidence · operations · schema	✅	✅	✅	✅	✅
One provider's Kin Score + agent-readiness score	✅	✅	✅	✅	✅
Tag-group and taxonomy membership (with bands)	✅	✅	✅	✅	✅
Scoring rubric — the methodology behind every published score	—	—	✅	✅	✅
Resolve · enrich — the identifier and field-group ergonomics	—	—	✅	✅	✅
Cohorts · capabilities · insights discovery	—	—	✅	✅	✅
Operation search — which APIs expose a given path	—	—	✅	✅	✅
Deprecation roll-up — what the catalog is retiring	—	—	✅	✅	✅
Venture capital — firms and portfolios	—	—	✅	✅	✅
Scores across a set (ranked lists, cohort rosters, tag rollups)	—	—	✅	✅	✅
Facets · trend · delta · regulatory on one provider	—	—	✅	✅	✅
Ratings leaderboard · movers · what changed	—	—	✅	✅	✅
Industries · Regions · Areas leaders	—	—	✅	✅	✅
Score history — a provider's series, movement from any date	—	—	✅	✅	✅
Stack design (recommend · blueprint · APIs.json export)	—	—	✅	✅	✅
OAuth scopes · security artifact collections	—	—	✅	✅	✅
Saved Workspace (saved searches · lists · net-new deltas)	—	—	✅	✅	✅
Agent-readiness leaderboard + dimension diffusion	—	—	✅	✅	✅
The investor graph on one provider	—	—	✅	✅	✅
Request a re-check of your listing (POST /checks)	—	—	—	—	✅
Compare · gap analysis — choosing a vendor, market whitespace	—	—	✅	✅	✅
Cohort stats · rankings · scores · capabilities — benchmarking a market	—	—	✅	✅	✅
Cohort history — how a market's scores moved over time	—	—	✅	✅	✅
Cohort failure modes — the checks a market most commonly fails	—	—	✅	✅	✅
Insights depth (full company profile · gaps · provider-match)	—	—	✅	✅	✅
Your own listing: ranked remediation, band gates, fix projections	—	—	—	—	✅
Request missing artifacts — apis.json, agent card, MCP, Arazzo, rules, scopes	—	—	—	—	✅
Claim your listing and correct facts, submit artifacts, set visibility	—	—	—	—	✅
Dispute a finding with evidence we fetch for you	—	—	—	—	✅
CI output — SARIF / JUnit, so a slipping score fails your build	—	—	—	—	✅
Watch your listing for score and band changes	—	—	—	—	✅
Data licence — whole rosters, /export datasets, co-branding	—	—	—	—	✅
The whole time series — every provider's score history, licensed	—	—	—	—	✅

Access is Understanding, bought by the call. Same reach, same rate limit, no subscription: you put a card on file, hold a prepaid balance, and every paid call debits it — a cent for a single-entity answer, a nickel for one that composes across the catalog. Every price is published at /api/v1/prices and quoted in an x-price-usd header on the call that charged it, and x-credit-balance shows the balance it left, so nothing about the bill is discovered afterwards. /api/v1/prices is the authority: its status reads live.

Lists and searches are priced by what they return, not by the request. Ranked lists, cohort rosters, operation search and the other collection reads cost $0.005 per request plus $0.0002 per record returned — less past 100 records and again past 1,000 — capped at $0.50 however wide the answer. A precise question is cheap; asking for a full page costs what a full page is worth. The charge follows the records you were handed, not how many matched, so paging is not a workaround and not a penalty. Ask what a call will cost before making it: /api/v1/prices/estimate?path=/v1/ratings&limit=50.

The balance tops itself up from your card when it runs low — $25 to start, stepping up if a top-up lasts under a week — and your spend caps always bind first: $50 a day and $500 a month unless you set others. Credit expires twelve months after your last top-up and is refundable on request before then.

It is deliberately not a cheaper Influence. Bulk export is priced per row and cannot be quoted before a call runs, so it stays on Influence.

Human work is bought one request at a time. A re-check of your listing ($99), a dispute, a facts correction, an artifact submitted or generated ($49 each) is a person doing work, so it is never debited from a balance and never needs a plan: a refusal on those paths links a one-off Checkout for exactly that request, with its own receipt. Influence includes them. And if we simply have your company wrong, report_correction or POST /api/v1/providers/{slug}/correction is free and needs no key at all.

Past roughly 20,000 calls a month, Understanding is cheaper — 3,980 if they are all cross-catalog. We would rather tell you that than take the difference, so we email you when your spend crosses $199.

The Explore row is the usage plan attached to an agent principal's own key. A caller sending no key at all gets the same access as Explore but is not metered by a plan — it is limited per IP at the edge. See rate limits for why, and what a key buys you before it buys volume.

The names on the wire

The plans are named for what you are doing; the API still identifies them by their original short names, and will keep doing so — they are in every 401 and 402 body, in X-RateLimit-Tier, and in the x-tier extension on every operation in the reference. If you are reading a response rather than this page, this is the mapping you need:

Plan	tier on the wire
Explore	free
Learn	starter
Understanding	pro
Influence	business

Understanding is per-query intelligence. Influence is the whole dataset in one pull. It pages through a large cohort roster (2,500 members per request, which covers all but five cohorts in a single call); Influence gets it whole, plus /v1/export/ratings, /v1/export/providers and /v1/export/history for licensing and republication.

Something wrong with your listing?

Reporting an error is always free. Use the report_correction MCP tool, or open an issue on your company's repo at github.com/api-evangelist/<your-slug> — that reaches a person, it is public, and you can see what we did with it. Neither needs a key or a plan.

Requesting a re-check is Influence. POST /v1/checks (or the request_check MCP tool) queues the work itself: re-profiling the target and running the full enrichment pipeline against its live surface, which a person runs by hand. You get an id back immediately, GET /v1/checks/{id} tells you where it stands, and if you leave an email we notify you when it changes. Getting a company profiled that is not listed yet works the same way.

A check targets a provider by default, or a slice of the catalog — pass target_type as industry, tag, area, or estate alongside the slug. That covers the things that go wrong one level up: an industry roster missing an obvious member, a tag that has drifted, an area whose ranking looks wrong.

A slug we do not already hold is a request to add it, not an error. That is true for every target type. A company nobody has catalogued, an industry the catalog is missing, an area worth carving out, an estate nobody here has thought of — all of them queue as an add request rather than bouncing. If the slug looks like a near miss you also get did_you_mean back with the request, so a typo is visible without anything being dropped.

What can we fix?

Everything above answers questions about the catalog. Influence also answers one about your own API: which specific checks you fail, what each is worth, in what order, and where your score and band land if you fix them. It separates a points gap from a band gate — some bands are not reachable by collecting points at all — and projects the result before you commit engineering time. The free improve_my_score still tells you which facet is weakest; this tells you which check to fix and what you get.

The time series

APIs.io records every provider's Kin Score and agent-readiness score on every scoring build. Understanding gets one provider's series and can ask what moved since any date; Influence licenses the whole thing. It is the one dataset here that cannot be reconstructed later — these are the only records of what the catalog looked like on those dates, and the set grows with every build.

Log in with GitHub   — get a free Learn key, then upgrade any time from your account.

How it works

1. Log in with GitHub — you're issued a free Learn API key instantly. Learn exists to put a name to your traffic, not to be a larger free plan: it raises the anonymous quota from 500 to 2,000 a day and the rate from 5 to 20 req/s, and everything beyond that is Understanding.
2. Send it as an X-API-Key header:

   Code
   curl -H "X-API-Key: <your key>" "https://apis.io/api/v1/industries"

3. Upgrade to Understanding or Influence from your account page (Stripe Checkout). Your key keeps working — its limits and access just increase immediately.

Without a key you get Explore access — the whole catalog, as above — but no usage plan: an anonymous caller is limited per IP at the edge rather than metered per key, because a plan meters a key and every anonymous caller would have to share one. The apis.io website runs on its own rate-limit tier and buys no paid data — it is identified by request origin, and a request origin is something any caller can claim, so it is never a credential. See Authentication and Rate Limits for details.

Every response that withholds a number says so in place, names the tier that returns it, and links here — so a client can act on the upgrade instead of guessing why a field is missing.

What a plan does not buy

A plan buys access to the catalog. It does not move a Kin Score, and it cannot.

Only artifacts a provider publishes themselves are scored. Buying any plan, at any price, sponsoring any placement, or applying any vendor profile changes no facet, check, band or ranking. This is not a policy we intend to keep — it is a property of how the scorer is built:

• The scorer cannot see who paid. There is no sponsorship, advertiser, vendor, paid-tier or billing field anywhere in the data it reads — not fields it declines to use, fields that do not exist in its input.
• No published check can be satisfied by a commercial relationship. Every check names a public artifact and the rule that reads it, and the full list is published so the claim can be audited rather than believed.
• A simulation is not a score. Where the catalog models what a rating would be, the figure is structurally distinguishable from the stored one and never replaces it.

The cheapest way to raise a Kin Score is still the free one: publish the artifact.

The full statement, and how to check each claim yourself, is at apis.io/rating/.

Every endpoint, by tier

The table above is the feature view. This is the path view — every route the API exposes, against the tier that unlocks it. Paths are shown relative to https://apis.io/api/v1/, and each tier includes everything above it.

Explore · Learn

Explore is the catalog itself: finding a company and reading what it publishes. Learn and Explore reach exactly the same paths — Learn buys identity and headroom (four times the quota and four times the rate), not additional endpoints.

• GET /search — the federated search across APIs, providers and tags
• GET /providers · /providers/{slug}
• GET /providers/{slug}/ → apis · artifacts · similar · onboarding · operations · tools · schema · capabilities · evidence · agent-readiness
• GET /apis · /apis/{aid} · /apis/{aid}/artifacts · /apis/{aid}/similar
• GET /tags · /tags/{slug} · /tag-groups · /tag-groups/{slug} · /tag-groups/{slug}/tags
• GET /industries · /regions · /areas — browse and get (the /leaders view under each needs Understanding)
• GET /openapis/{aid} — one API's primary OpenAPI reference, ?include=content to inline it
• GET /playground — curated APIs that need no key and cost nothing, each with a working example request. Where you start if you are new here, which is why it is not behind a plan

Artifact collections: GET /openapis · /asyncapis · /arazzo · /postman · /collections · /graphql · /json-schemas · /json-structures · /json-ld · /rules · /examples · /finops · /plans · /rate-limits · /apis-json · /mcp · /skills · /channels

/scopes and /security are the two collections that are not free — see Understanding.

Everything about ONE provider stays free, including its headline Kin Score and agent-readiness score. What a free key cannot do is read those numbers across a set: any result carrying more than one record returns bands and drops composite and the agent-readiness number. That holds however you ask — the default projection, fields=composite, and view=full all redact.

Understanding

Everything that synthesises over the catalog rather than reporting one record of it.

• GET / — the service root and catalog counts
• GET /resolve · /enrich
• GET /ratings · /ratings/movers — the leaderboard and who is rising or falling, and /ratings/rubric, the scoring methodology behind every published score
• GET /changes — what moved in the catalog since a date
• GET /operations — catalog-wide operation search: which providers expose a path matching your terms, e.g. ?q=refunds&method=POST. A tag is what a company says about itself; a path is what its API accepts
• GET /operations/deprecated — every provider retiring operations, ranked by how many
• GET /scopes · GET /security — the two paid artifact collections
• GET /industries/{slug}/leaders · /regions/{slug}/leaders · /areas/{slug}/leaders
• GET /providers/{slug}/rating · /rating/facets · /rating/history · /business-capabilities
• GET /capabilities · /capabilities/{id} · /capabilities/{id}/edges — the business-capability model: what these APIs let a business do, with the quoted contract fragment behind each claim
• GET /providers/{slug}/business-capabilities — the same question for one vendor
• GET /cohorts · /cohorts/{kind}/{slug} — the roster (its analysis parts need Influence)
• GET /insights · /dimensions · /adoption · /industries · /companies · /insights/company/{slug} at preview depth
• GET /stack · /stack/export — the Stack Builder and its APIs.json export
• GET /vcs · /vcs/{slug} · /vcs/{slug}/portfolio — the venture layer
• GET /providers/{slug}/investors — which firms back one company
• GET /compare — two providers side by side on one rubric
• GET /gaps · /gaps/industry/{slug} — where a provider, a stack or a whole industry is thin
• GET /agent-readiness · /agent-readiness/dimensions — the catalog-wide agent leaderboard, and how far each dimension has actually diffused
• GET /cohorts/compare and /cohorts/{kind}/{slug}/stats · /rankings · /scores · /capabilities · /history · /failures — a market's distribution, leaderboard, movement over time, and the checks it most commonly fails
• GET /insights/company/{slug} at full depth · /gaps · /match
• GET /me — the workspace root: your plan, what you have saved, and where it lives
• /me/* — the Saved Workspace: GET|POST|DELETE /me/searches, /me/lists, and /me/searches/{id}/results, /net_new, /entries

Understanding also lifts the redaction above: cohort rosters, tag-group rosters and their stats, ranked provider lists, and readiness_total on /insights/companies all return real numbers.

Influence

Your own listing. Everything else on this page is about reading the catalog. Influence is about the entry in it that belongs to you — what it says, what it scores, and changing both.

Know what to fix

• GET /providers/{slug}/remediation — every rubric check you fail, what each is worth, and the order to fix them in. Add ?format=sarif or ?format=junit to fail your own build on it
• GET /providers/{slug}/gates — which bands are closed to you regardless of score
• GET|POST /providers/{slug}/projection — where you land if you make named fixes

Ask for what you are missing

• POST /providers/{slug}/generate — what we can make for you
• POST /providers/{slug}/generate/{artifact} — request an artifact we mark you down for not having: apis-json, agent-card, mcp (one tool per operation you already publish), arazzo, rules (a Spectral ruleset scoped to your own failing checks), and scopes / security read out of your live OpenAPI. A person generates it, checks it, and sends it to you
• Nothing is published on your behalf. Your score moves when you commit the file

Claim it, then change it

• POST /providers/{slug}/claim — returns a token and the hosts it will be checked against. Publish it as a DNS TXT record or a file at /.well-known/apis-io-challenge, and leave it up. A person verifies it and emails you; there is nothing else to call
• POST /providers/{slug}/submit — point us at an artifact you publish
• POST /providers/{slug}/facts — propose corrections to your name, description, url, industries or tags
• POST /providers/{slug}/visibility — ask to be restricted (name, description, your own link, unrated) or delisted entirely

Tell us we are wrong, and keep watch

• POST /providers/{slug}/dispute — "you say we lack X, here it is." Send the claim and a URL; a person fetches your evidence and emails you either way
• POST /checks — ask a person to re-profile you, or to add something we do not hold at all
• GET /checks/{id} · GET /checks/mine — where every request you have filed stands
• POST|DELETE /me/watch/{slug} — be told when your score or band moves; GET /me/watch lists them

Every request is worked by a person

All of the above return 202 Accepted with an id and a status URL. Nothing is applied, verified or generated on the call, and the response says so: "completed": false.

That is deliberate. A check means re-profiling you and re-running the enrichment pipeline against your live surface; a claim means someone confirming a record you published is really yours; a generated artifact is one we have looked at before sending. None of those are things an API can honestly do in the time it takes to answer a request — and a 200 would tell you otherwise.

Poll GET /checks/{id}, or leave a contact and we email you when the status changes.

Two things that stay free, and always will: reporting an error (report_correction, or an issue on your own repo at github.com/api-evangelist/<your-slug>), and everything anyone can read about you. Influence buys the work — a person re-profiling you — and the tooling to do it yourself.

Changing what we hold needs a verified claim. Reading your own remediation does not. Without that line, any paying caller could edit or hide a competitor. Requests that change a listing are queued for a person rather than applied on the call: delisting strips artifacts, pages and rollups across the network, which is not an irreversible change to make on one API request.

Also sold here: the data licence

GET /export/ratings · /export/providers · /export/history, with ?co_brand= for licensees, plus cohort rosters whole rather than paged. This is not listing management — it is the whole scored dataset in one pull, for an outlet or analyst republishing the numbers. It sits on this plan because it is the only paid line above per-query use, not because it belongs with the rest.

How a gate answers

A gated path answers 401 Unauthorized when no key was sent and 402 Payment Required when a valid key is below the required tier — never 403. Both bodies name what you asked for, which plan returns it, and what the single call would cost on the pay-as-you-go plan. The sentence uses the plan names; the tier field carries the wire name, so a client can branch on it without parsing English. The 402:

Code
{
  "error": "upgrade_required",
  "detail": "The ratings endpoint requires the Understanding, Influence or Access plan.",
  "tier": "pro",
  "plan": "Understanding",
  "plans": "https://apis.io/developer/plans/",
  "price_usd": 0.025,
  "band": "PR",
  "access": {
    "plan": "Access",
    "detail": "Pay per call instead: this one costs at most $0.025, charged by the records returned. Add a card, hold a balance, and every call debits it — no subscription.",
    "get_started": "https://apis.io/developer/account/?access=1&from=ratings",
    "prices": "https://apis.io/api/v1/prices"
  }
}

The 401 carries the same tier, plans, price_usd and band, with error: "unauthorized" and a Bearer challenge naming the resource metadata, because a caller with no identity is being asked to get one before being asked to pay. On the wire the challenge arrives as x-auth-challenge (and x-amzn-remapped-www-authenticate), not under the standard WWW-Authenticate name — API Gateway renames that header and CloudFront cannot rename it back on a refusal — so read either of those two. The price is also on both responses as x-price-usd / x-price-band headers, for browser clients that cannot read the body of a failed cross-origin fetch.

pro names the lowest plan that unlocks the path, and the sentence says "Understanding or Influence" because either one does — being told to buy Understanding while already holding Influence would read as a refusal of the plan you are on — and it names Access too, because a card and a balance also unlock it. A list call quotes price_usd at the most it can cost for the limit you asked (here, a 100-record page), which is also what your balance must cover before it runs; you are charged for the records actually returned.

401 and 402 are deliberate: 403 and 404 are rewritten to the site's error page before they reach you, so the explanation would arrive as HTML your client cannot parse. The one exception is a resource restricted to the catalog operator (story-leads, the admin surface), which answers 402 with error: "not_purchasable" — no plan unlocks it, and the code is chosen to survive the rewrite rather than to invite a payment.

This page groups the surface by what you would use it for. The API reference is the complete list — every one of the 100 routes, each declaring its plan as x-tier, and each one checked against what the gate actually enforces rather than what we meant to build. Where a response is shaped rather than refused, the withheld field is replaced in place by a gated object carrying the same tier and link, so a client can act on it without a second request.

Every MCP tool proxies one of these paths, so the boundary is identical on both surfaces, and the tool descriptions name the same plans — a tool marked UNDERSTANDING is the Understanding plan. The API reference declares the plan for every operation as x-tier, using the wire name.