How Much Does a Bitcoin API Cost? Prices Checked August 2026

Pricing a Bitcoin data vendor is harder than it should be, and the reason is that almost nobody charges for the same thing. One provider sells you requests. Another sells you points that vary with how you wrote the query. Google sells you bytes scanned. Running your own node sells you nothing at all and then bills you for a server every month until you turn it off.

So a comparison of monthly prices alone is close to meaningless. What follows is the sticker price for each of the main options, checked on 9 August 2026, plus the thing that actually determines your bill: how each one counts.

What a Bitcoin API costs in August 2026

Every figure below came from the provider's own pricing page or documentation on 9 August 2026. Vendor pricing changes without much announcement, so treat this as a dated snapshot and confirm before you sign anything.

Two entries in that table deserve a caveat rather than a number. Blockchair publishes API plans but its documentation pages return HTTP 401 to an unauthenticated fetch, and when we sent a plain two-address request to its dashboards endpoint in an earlier test the response was code 430 with the message that the IP address is temporarily blacklisted and to apply for an API key. That is a rate limit rather than a price, but it tells you the same thing: the free path is not a production path. And mempool.space genuinely does run an open API with no key, which is excellent of them and not something to build a commercial product on top of without an arrangement.

The four billing models, and which one bites

Sticker price is the part everyone compares. The billing unit is the part that determines whether the sticker price holds.

Per request. The easiest to forecast. You know roughly how many address lookups your app makes per day, you multiply, you have a number. BlockCypher's tiers work this way. The failure mode is a burst: a rate limit expressed per second means a batch job can breach it while your monthly average sits comfortably inside the plan.

Per credit or per point. Bitquery meters what it calls API points, and its documentation defines them as resources consumed multiplied by a price per unit, with consumption rising according to the number of records queried and the complexity of the GraphQL operation. Their realtime dataset is a flat 5 points per cube. The consequence is that two engineers writing the same question in different shapes get different bills, and you cannot compute your cost from your traffic alone. One detail in their docs is worth knowing before you build retry logic: points are deducted even when your client times out, because points reflect work the backend performed. A badly tuned retry loop can spend real money producing no data.

Per byte scanned. This is the BigQuery model, and it is the one that surprises people hardest. The bill follows the columns and partitions your query touches, not the rows it gives back. A query returning a single balance can scan hundreds of gigabytes if it aggregates every output ever paid to an address, and it costs the same every time you run it. Partitioning and dry runs help, but exploration is exactly the phase where you do not yet know which columns you need. We wrote up that trap in more detail in our guide to the BigQuery Bitcoin dataset and what it costs.

Per server. Running Bitcoin Core yourself removes the vendor entirely and replaces it with infrastructure you own. The software is free. The machine is not, and neither is the several hundred gigabytes of storage or the bandwidth if you leave uploads unrestricted. We costed this out separately in how much it costs to run a Bitcoin node. The number that catches teams out is not the monthly server bill but the engineering time: a node gives you blocks and transactions, not balances, labels or history, so you are still going to build an indexer.

How much does a Bitcoin API cost for a small project?

For a hobby project or a prototype, close to nothing. Public endpoints from mempool.space and Blockstream serve address and transaction lookups with no key at all, and BlockCypher's free tier covers 100 requests an hour, which is plenty for a dashboard you refresh by hand. If your usage is a few thousand calls a month and you can tolerate an occasional HTTP 429, you should not be paying anyone.

The line gets crossed at the point where a failed request becomes a customer-visible bug. That is usually when a real product starts, and it arrives sooner than people expect, because 3 requests per second is one moderately busy page.

Why is Bitcoin historical data more expensive than current data?

Because current data is small and historical data is not. The unspent output set is a few gigabytes and every node keeps it. The full history of every movement of every address since 2009 is a much larger structure that somebody has to build, store and keep in sync, and it is the thing free explorers almost universally do not offer. Ask any public explorer what an address held on a past date and you will get its balance today, because that is the only version it stores.

That gap is exactly where paid providers earn their money, and it is worth checking carefully during evaluation. Bitquery, for instance, does support point-in-time balances by filtering inputs and outputs with a height condition, though its own documentation notes that the pre-aggregated addressStats view may be stale and advises computing the exact figure from inputs and outputs. We walked through the whole problem, including four public endpoints tested against one address, in how to check a Bitcoin address balance on a specific date.

Is there a free Bitcoin API?

Yes, several, with real limits attached. mempool.space and Blockstream both run open Esplora-style APIs that need no key. BlockCypher gives you a free tier of 3 requests per second and 100 per hour. Google's BigQuery public dataset includes the first terabyte of scanning each month at no charge.

What none of them give you is a commercial guarantee. Free tiers are throttled, unversioned in practice, and can be withdrawn. Bitquery's cheapest paid plan at $39 per month is explicitly marked personal use only on its pricing page, which means a US company evaluating it commercially starts at the $79 Pro tier, not the advertised entry price. That kind of licensing detail is easy to miss and expensive to discover after you have shipped.

How to estimate your own bill before you commit

Work forward from traffic rather than backward from a plan. Four numbers get you most of the way.

Start with calls per day at steady state, then add your peak burst per second, because that is what trips rate limits rather than monthly totals. Add the share of those calls that need history rather than current state, since that is what pushes you into a higher tier. Finally, add the number of distinct addresses you touch per request, because batch endpoints and per-address endpoints price very differently.

Then read the plan against those four numbers instead of against its headline. A $79 plan with a 90 request per minute ceiling is not comparable to a $119 plan measured per second until you convert both to the same unit. If your data vendors sit alongside a growing cloud bill, it is worth keeping a read-only view of what your cloud and SaaS spend actually is month to month, because API costs tend to creep quietly between reviews rather than jump.

What we charge and why the unit matters

BitcoinDatabase indexes the Bitcoin chain once and serves it three ways: a REST API, SQL and dashboards. Pricing is usage-based on calls and queries, and there is no free plan, which is a genuine disadvantage if you want to evaluate at zero cost first. The tradeoff is that your bill follows the number of questions you ask rather than the shape in which you asked them, and balances, UTXOs, entity labels and fund flows are already indexed, so a historical balance is a lookup rather than an aggregation you pay to recompute.

If you are weighing us directly against the multi-chain option, we put the numbers side by side on our Bitquery alternative page, including where Bitquery is the better choice. For the wider field, our roundup of the best Bitcoin API for developers covers capability rather than price, and the Bitcoin API page shows what the endpoints actually return.

The costs that do not appear on any pricing page

Three of them, consistently.

The first is reconciliation. When two providers disagree about an address, and they do, somebody on your team spends a week working out which is right. We have seen two explorer families differ by roughly 50 BTC on a single well-known address for a documented reason, and that investigation is not free.

The second is the aggregation layer. Raw chain data is cheap and answers almost no business question directly. Bitcoin stores no balances, so every balance is computed, and if your provider does not compute it, you will, forever, on every query.

The third is migration. Billing units differ enough between providers that switching means rewriting your query patterns, not just your API key. That is the strongest argument for spending an afternoon on the estimate before you pick, rather than a quarter on the move after you have.

All prices and limits above were checked on 9 August 2026 against each provider's own pricing page or documentation. Provider pricing changes, so verify current figures before budgeting. This article is informational on-chain data and tooling guidance, not investment, financial or legal advice.