API · Reference

What One Credit Buys, Endpoint by Endpoint

CoinMarketCap APIUpdated 14 September 2026 · 11 min read
What One Credit Buys, Endpoint by Endpoint, shown as a halftone API wordmark above a layered code-window stack on cobalt.

A CoinMarketCap plan is priced in call credits, not calls. Two integrations making the same number of requests can differ tenfold in cost, and the difference is almost never the request count.

This page lists what every endpoint charges, including the thirty that do not publish a figure, and works through what that comes to at realistic volumes.

Key Takeaways

  • There are three billing models, not one: per batch of results returned, one flat credit per call, and one credit per item returned.
  • Most current endpoints bill per 250 results, rounded up. The general guide’s figure of 100 now describes only the historical endpoints.
  • The DEX family bills one flat credit per call, whatever the page size. The batch rule does not apply to it.
  • Cadence costs more than breadth. Scanning 5,000 assets hourly is cheaper than polling 100 assets every minute.
  • Ten endpoints are free, including the ID map, usage stats, and everything under /v1/content/ and /v1/community/.

The Three Models

Every priced endpoint uses one of three shapes.

  • Per batch1 credit / 250 results

    The endpoint declares a batch size and charges one credit per batch of results returned, rounded up. A call returning 300 results on a 250-batch endpoint costs 2 credits. This covers most market-data endpoints.

  • Flat per call1 credit / call

    One credit, whatever comes back. This covers the indices, the airdrop endpoints, the fiat map, the CMC AI market feed and coverage map, and the entire DEX family.

  • Per item1 credit / cryptocurrency

    One credit per item returned. Only /v5/cmc-ai/coins/latest bills this way, at one credit per cryptocurrency.

convert is a multiplier

Each convert currency beyond the first adds a credit. Requesting three fiat conversions triples the cost of a call, which is the single easiest way to overspend without noticing.

Error responses are never charged. Measured across two probe rounds, every non-200 response returned credit_count 0.

A Third of the Surface Publishes No Cost at All

Re-censused on 14 September 2026 across every endpoint reference page:

74Endpoint reference pages
44Pages stating a “Plan credit use”
30Pages stating no cost

Two things about that 30 are worth knowing before you plan a budget. Almost all of it is the DEX family, which is priced in practice but documented nowhere, and the count has grown: the same census on 11 September found 66 pages and 22 unpriced, so every endpoint added since then shipped without a cost line.

The machine-readable corpus carries none of it

The gap is wider than the reference pages suggest. llms-full.txt, which CoinMarketCap’s own documentation describes as a single-file dump of all documentation, contains the string “Plan credit use” zero times in 263,601 characters, because it carries the guides and overviews but none of the endpoint reference pages. So an agent that ingests the documentation the way the documentation tells it to will not find a single credit cost.

Every Endpoint, and What It Costs

Costs marked measured are not published by CoinMarketCap. They were obtained by calling the endpoint and reading credit_count, on 11 September 2026.

Cryptocurrency

Endpoint Charged on Cost
/v3/cryptocurrency/listings/latest cryptocurrencies returned 1 per 250
/v3/cryptocurrency/quotes/latest cryptocurrencies returned 1 per 250
/v2/cryptocurrency/info cryptocurrencies returned 1 per 250, no convert charge
/v2/simple/price cryptocurrencies returned 1 per 250
/v1/cryptocurrency/category cryptocurrencies returned 1 per 250
/v1/cryptocurrency/listings/new cryptocurrencies returned 1 per 250
/v1/cryptocurrency/trending/latest cryptocurrencies returned 1 per 250
/v1/cryptocurrency/trending/gainers-losers cryptocurrencies returned 1 per 250
/v1/cryptocurrency/trending/most-visited cryptocurrencies returned 1 per 250
/v2/cryptocurrency/market-pairs/latest market pairs returned 1 per 250
/v2/cryptocurrency/price-performance-stats/latest cryptocurrencies returned 1 per 250
/v1/cryptocurrency/categories request plus cryptocurrencies 1 flat plus 1 per 200
/v3/cryptocurrency/quotes/historical data points returned 1 per 100
/v1/cryptocurrency/listings/historical cryptocurrencies returned 1 per 100
/v2/cryptocurrency/ohlcv/historical OHLCV data points returned 1 per 100
/v1/cryptocurrency/airdrops request 1 flat
/v1/cryptocurrency/airdrop request 1 flat
/v2/cryptocurrency/ohlcv/latest not published 1measured
/v1/cryptocurrency/map not published 0measured

/v1/cryptocurrency/categories is the only endpoint with a two-part charge, and the only one still on a 200 batch. Budget it separately.

Exchange, Indices and Global Metrics

Endpoint Charged on Cost
/v1/exchange/listings/latest exchanges returned 1 per 250
/v1/exchange/quotes/latest exchanges returned 1 per 250
/v1/exchange/info exchanges returned 1 per 250, no convert charge
/v1/exchange/market-pairs/latest market pairs returned 1 per 250
/v1/exchange/quotes/historical data points returned 1 per 100
/v1/global-metrics/quotes/historical data points returned 1 per 100
/v1/exchange/map call 1 flat
/v1/global-metrics/quotes/latest call 1 flat
/v3/fear-and-greed/latest, /v3/fear-and-greed/historical request 1 flat each
/v1/altcoin-season-index/latest, /v1/altcoin-season-index/historical request 1 flat each
/v3/index/cmc100-latest, /v3/index/cmc100-historical call 1 flat each
/v3/index/cmc20-latest, /v3/index/cmc20-historical call 1 flat each
/v1/exchange/assets not published 1measured

The indices charge a flat credit however much they return. A full historical Fear and Greed series costs the same single credit as the latest reading, which makes backfilling one of them effectively free.

DEX and On-Chain

Every endpoint in this family bills one flat credit per call, and none of them publishes that. Page size makes no difference: /v1/dex/platform/list and /v1/dex/search each returned 1 credit at limit=1 and at limit=100.

Endpoint Cost
/v1/dex/search 1measured
/v1/dex/token, /v1/dex/token/price, /v1/dex/token/pools 1 eachmeasured
/v1/dex/tokens/transactions 1measured
/v1/dex/liquidity-change/list 1measured
/v1/dex/holders/count, /v1/dex/holders/tag_count 1 eachmeasured
/v1/dex/platform/list, /v1/dex/platform/detail 1 eachmeasured
/v4/dex/pairs/quotes/latest 1measured
/v1/k-line/points, /v1/k-line/candles 1 eachmeasured
/v1/dex/token-liquidity/query, /v1/dex/security/detail, /v4/dex/spot-pairs/latest, /v1/dex/holders/trend/list Not published, not yet priced

For a DEX integration the arithmetic is therefore simple: count your calls. Page size is free, so request as much per call as the endpoint allows.

CMC AI

Endpoint Charged on Cost
/v5/cmc-ai/latest request 1 flat
/v5/cmc-ai/coins/map request 1 flat
/v5/cmc-ai/coins/latest cryptocurrencies returned 1 per coin

Three exemptions apply on the per-coin endpoint, and they are the reason it costs less in practice than the headline rate suggests. A coin returned with an empty insights[] is not charged: if CoinMarketCap has generated nothing for that asset at that moment, the coin comes back in the response and costs nothing. A coin skipped by skip_invalid is not charged, so unresolvable identifiers in a batch are dropped for free. And a repeat call returning the same content is not charged, so because answers regenerate on a schedule between 30 minutes and 24 hours, polling more often than the content changes does not multiply the bill.

That last one is unusual enough to plan around: a conservative polling schedule against these endpoints costs materially less than its call count implies.

Free

Endpoint Cost
/v1/content/latest, /posts/latest, /posts/top, /posts/comments 0
/v1/community/trending/token, /v1/community/trending/topic 0
/v1/cryptocurrency/map 0measured
/v1/key/info 0measured
/v1/tools/postman Returns a Postman collection, no credit_count

Ten endpoints cost nothing. Two of those matter operationally: the ID map means refreshing your identifier table is free, and /v1/key/info means you can poll your own usage as often as you like without the monitoring itself consuming budget.

What It Comes to in Practice

Four builds, costed against the rules above. Assume a 30-day month.

Build Pattern Credits per call Per month
Top-100 price dashboard quotes/latest, 100 coins, every minute 1 43,200
Whole-market scan listings/latest, 5,000 assets, hourly 20 14,400
Same scan, badly paged listings/latest, 50 calls of 100, hourly 1 × 50 36,000
Top-100 dashboard, three fiats as above with convert=USD,EUR,GBP 3 129,600

Arithmetic derived from the per-endpoint costs above, on a 30-day month.

Three things fall out of that table.

  1. Cadence dominates breadth.

    Scanning the entire market hourly costs a third of what polling 100 coins every minute costs. Before optimising page size, ask whether the interval is right: on Basic and Builder the data only refreshes every 60 seconds anyway, so a faster poll returns identical values and pays for them.

  2. Paging below the batch is the expensive mistake.

    The same 5,000 assets cost 14,400 credits a month at limit=5000 and 36,000 at limit=100. Identical data, two and a half times the price, and fifty times the calls against your per-minute limit. The ceiling is 5,000 per call on the listings endpoint.

  3. convert is a multiplier, not an add-on.

    Three currencies triple the bill. If you need several fiats, consider converting once against a stored rate rather than asking the API to do it per call.

A CMC AI example, for contrast: a portfolio tracker with 20 holdings, refreshed three times a day against the 8-hour regeneration cycle, costs at most 20 credits per call and 1,800 a month. In practice it costs less, because holdings with no generated content are free and repeat calls returning unchanged answers are not charged.

Sizing Against a Plan

Tier Monthly credits Rate limit Update frequency
Basic (free) 15,000 50 req/min 60 seconds
Builder 150,000 300 req/min 60 seconds
Startup 450,000 600 req/min Real-time
Growth 2,000,000 750 req/min Real-time
Professional 5,000,000 1,200 req/min Real-time
Enterprise 30,000,000 1,600+ req/min Real-time

Figures from the CoinMarketCap API pricing page, verified 11 September 2026.

Read the credit column and the rate-limit column as separate constraints, because they bind differently. The badly paged market scan above uses 36,000 credits a month, which fits comfortably inside Builder, but its 50 calls per scan against a 300-per-minute limit is fine while the same job at limit=100 on Basic would need careful pacing against 50 per minute.

Two operational notes. Credit windows on paid plans are anchored to your subscription date rather than the calendar, so a plan started on the 5th refreshes on the 5th. And email notification arrives at 95% of the monthly limit, which is usually later than a product wants to find out.

Checking Rather Than Calculating

Every response reports what it actually cost.

python
import os
import requests

r = requests.get(
    "https://pro-api.coinmarketcap.com/v3/cryptocurrency/listings/latest",
    headers={"Accept": "application/json",
             "X-CMC_PRO_API_KEY": os.getenv("CMC_API_KEY")},
    params={"start": "1", "limit": "250", "convert": "USD"},
)
status = r.json()["status"]["credit_count"]
print(f"this call cost {status} credit(s)")

Log credit_count in development and you will find your expensive calls in an afternoon, without modelling anything. For the thirty endpoints that publish no cost, this is the only way to know.

/v1/key/info reports usage against your plan and costs nothing, so a burn-rate alert built on it is free to run.

Where to Go Next

For the credit model explained rather than tabulated, plus the four rate limits and how to set a polling interval, read the credits guide. WebSocket is billed separately at 0.025 credits per message received and sits outside this table.

FAQ

What does one credit buy?

It depends on the endpoint. On most current market-data endpoints, up to 250 results. On the historical endpoints, up to 100 data points. On the indices, the airdrops and the whole DEX family, one call of any size. On /v5/cmc-ai/coins/latest, one cryptocurrency.

Does a bigger page cost more?

On a batched endpoint, yes, in proportion: 500 results on a 250-batch endpoint is 2 credits. But a page smaller than the batch wastes money, because 100 results also costs 1 credit. On the DEX endpoints page size is free.

Which endpoints are free?

Ten: the four /v1/content/ endpoints, both /v1/community/trending/ endpoints, /v1/cryptocurrency/map and /v1/key/info (both measured rather than published). /v1/tools/postman returns a collection file with no credit accounting at all.

Am I charged for a coin you have no data for?

No. On /v5/cmc-ai/coins/latest, a coin returned with an empty insights[] array is free, and so is a coin skipped by skip_invalid. Repeat calls that return the same content for a coin are also not charged.

Are failed calls charged?

No. Every error response measured returned credit_count 0, including authentication failures, validation errors and server errors.

How much does the DEX API cost per call?

One credit, on every endpoint tested, regardless of how many results come back. Four DEX endpoints could not be priced and none of the family publishes a figure, so read credit_count to confirm for your own usage.

Does convert cost extra?

Yes, one credit per conversion beyond the first, on every endpoint that supports it except /v2/cryptocurrency/info and /v1/exchange/info. Three currencies means three times the cost.

How do I know what a call actually cost?

Read credit_count in the status object of the response. It is present on every call and it is authoritative, which matters because a third of the endpoint surface publishes no cost.


The platforms, protocols, and projects referenced in these articles are used for illustrative purposes only to demonstrate how the CoinMarketCap API can be integrated in practice. CoinMarketCap makes no representation that these projects endorse or are affiliated with CoinMarketCap. If you represent a referenced project and have concerns, please contact us at content@coinmarketcap.com and we will respond promptly.