NexusNEXUS
Compare Nexus and Koios for Cardano. A managed full-stack data API with an SLA versus a community-run, low-cost endpoint. Honest trade-offs for both.
Comparison

Nexus vs Koios: Managed Full-Stack API or Community-Run Endpoint?

Nexus or Koios for your Cardano data layer?

Full disclosure before anything else: we know Koios well because we use it. Koios is one of the fallback providers in Nexus's own Cardano read path, and Gero Labs appears on Koios's community page among the projects building on it. So this is not a takedown of a competitor; it is an engineering comparison between two deliberately different models, written by a team that runs against both. Koios is an open-source, DAO-governed, community-operated data layer that sets the price floor of the entire Cardano API market. Nexus is a managed commercial platform that covers the same on-chain layer and then a lot of surface Koios intentionally does not attempt.

The decision between them is rarely about quality. It is about which failure modes, cost structures, and scopes fit your product. Everything below was verified against primary sources in July 2026 and re-checked at the end of September 2026. Koios is in an actively good state: spec v1.4.2, shipped on 2026-07-14, is still the current release, and the repositories show commits as recently as September 2026.

The timing matters. Maestro announced on 18 August 2026 that its developer API retires on 18 September 2026, which leaves Blockfrost, Koios and Nexus as the practical shortlist for a hosted Cardano data layer. If you are arriving from Maestro, the Maestro alternatives guide covers the migration itself; this page is about the choice between two of the three.

TL;DR

Koios and Nexus overlap on chain data and diverge on everything else. Koios is the cheapest serious source of Cardano chain data, open source and community-run, with no SLA at any price and payment in ADA only. Nexus is a managed commercial API that covers the same chain data and adds market data, wallet P&L, transaction building, swap routing and streaming, with invoices and a contractual SLA on its Dedicated tier. Pick on scope and accountability, because on price per chain read Koios wins.

Koios

Nexus

Model

Open source, DAO-governed, community-operated

Managed commercial service

Entry price

$0 (5,000 requests/day, no key)

$9/month Starter; Builder $29/month with a 7-day trial

Paid tiers

~$29.99 and ~$74.99/month, paid in ADA

$79, $199, $499/month, card and invoice

Chain data

Yes, about 102 endpoints

Yes, about 70 endpoints

Conway governance

15 current endpoints, the broadest available

13 governance and DRep endpoints

Market data, wallet P&L

No

Yes, as add-ons

Transaction building

No, submission of signed CBOR only

Yes, 6 builders returning unsigned CBOR

Streaming

No

Yes, WebSocket

SLA

None, at any tier

Contractual on Dedicated

Other chains

Cardano only

Bitcoin, Midnight, Apex Fusion

Rule of thumb: if your workload is chain reads and cost is the constraint, use Koios. If your product needs prices, P&L, transaction construction, streaming, or a contract behind its data layer, that is what Nexus sells.

Where Nexus fits: one account and one base URL across Cardano, Bitcoin and Midnight, flat monthly pricing from $9, with a 7-day trial on Builder, and the layers above chain data that Koios deliberately leaves out.

Two architectures, honestly described

Koios is a federation. Each instance runs the guild-operators stack: cardano-node, cardano-db-sync into PostgreSQL, Koios's own SQL normalization and cache layer (all MIT-licensed, every query public), PostgREST to expose it, and HAProxy in front. Independent community operators (the project commits to at least four trusted mainnet instances; Eternl, NMKR, VESPR, and TosiDrop run their own) register into a monitoring and proxy layer that does DNS-based load balancing, health checks that verify instances are serving tip-synchronized data, DDoS protection, and tier enforcement via JWT. There are at least two DNS entrypoints to avoid a single point of failure, and you can bypass the cluster entirely and hit an instance directly. Governance is a DAO grown out of the Guild Operators group, with five named core maintainers.

Two engineering consequences of that design worth knowing before you build on it. First, reads are served by whichever healthy instance the proxy picks, so Koios documents that paginated queries should always pass an explicit order parameter, or sort can differ between pages served by different backends. Second, accountability is structural, not contractual: health checks and failover are real and well-built, but nobody owes you anything (more on the terms below).

Nexus is a managed service: a Java 21 / Spring Boot API in front of a prioritized provider chain per network, with Nexus's self-hosted yaci-store indexers as the primary read path and Koios and Blockfrost as circuit-breakered fallbacks. Health checking includes a tip-staleness test that routes around an indexer that is alive but serving stale data. Above the chain-data layer sit the in-house market-data indexer, the analytics engine, the transaction builder, the DEX aggregator, and the WebSocket broadcaster. One operator, one bill, one accountable counterparty.

Koios's model gives you transparency and exit; you can setup-grest.sh a bit-identical private instance and your client code does not change. Nexus's model gives you accountability and scope. Those are genuinely different virtues.

API surface

Koios (api.koios.rest, base https://api.koios.rest/api/v1, plus preprod and preview hosts) documents about 102 endpoints (roughly 91 once you exclude the 11 deprecated-but-still-served ones): network and epoch state, blocks, transactions (including CBOR fetch and submission of signed transactions), addresses and payment credentials, stake accounts (13 endpoints), assets (13), pools (a deep group of 16, down to per-epoch delegator history), scripts and datums, an Ogmios HTTP passthrough, and a genuinely excellent Conway governance group: 15 current endpoints (18 counting deprecated ones) covering DReps (list, info, metadata, updates, voting-power history, delegators), the committee, proposals, per-proposal voting summaries, vote lists, and pool voting-power history, with correctness fixes as recent as v1.4.2.

The query model is PostgREST, and for batch-heavy consumers it is a real feature: vertical filtering with ?select=, horizontal filtering with PostgreSQL operators (?pool=eq.pool1...), CSV output via the Accept header, offset/limit pagination with Content-Range (1,000 rows per page), and bulk POST bodies (send 25 tx hashes in one call). Koios itself calls bulk queries its major differentiator, and it is right. The cap to know: request bodies are limited to 1 KB on the public tier and 5 KB registered, and blowing it returns a 413 (documented as of v1.4.2 after users hit it at around 100 IDs per call).

What Koios deliberately does not have, confirmed against the full spec: no market, price, or DEX data of any kind; no P&L or portfolio analytics; no transaction building (submission of signed CBOR only); no WebSockets, webhooks, or streaming. Koios competes on chain data and price, not breadth. That is a design choice, not a failure.

Nexus covers the same on-chain layer with about 70 Cardano chain-data endpoints (accounts, addresses, assets, blocks, epochs, pools, scripts, datums, submit, evaluate, batch CBOR) and then the layers above it: roughly 54 market-data endpoints (current and historical prices, OHLCV candles at 15m/1h/1D/1W, RSI(14), real and AMM-simulated order books, pool TVL and reserves, recent swaps, top traders, NFT floors and sales) aggregated across venues spanning 15 routable DEX integrations (live-verified September 2026: Minswap V1/V2, SundaeSwap V1/V3, WingRiders V1/V2, VyFi, Splash, TeddySwap, MuesliSwap, CSwap, Genius Yield, SaturnSwap, ShadowBook, SnekFun); FIFO wallet P&L including multi-wallet aggregation in one POST; a 6-operation transaction builder returning unsigned CBOR (transfers, delegation, stake registration with optional reward withdrawal, DRep registration, vote delegation with Trezor-compatible certificates, withdrawals) signed client-side via CIP-30; a DEX aggregator from quote to co-signed submission at a 0.10% fee (0.05% for GERO holders); SockJS/STOMP WebSocket streaming with snapshot-then-update semantics; 13 governance and DRep endpoints; a hosted MCP server with 25 tools; and live Bitcoin (including ordinals and runes), Midnight (including DUST and transaction building), and Apex Fusion networks, with keys scoped per chain and network.

On governance specifically, credit where due: Koios's governance group is still broader than Nexus's today. If your product is a governance explorer, Koios is arguably the best free source on the market.

Tiers, limits, and the fine print

Koios's tiers, as we last verified them in July 2026. Check the live page before you budget against it:

Public

Free

Pro

Premium

Custom

Price

$0, no key

$0, key required

~$29.99/mo

~$74.99/mo

negotiable

Requests/day

5,000

50,000

500,000

1,200,000

"unlimited"

Rate limit

100 per 10s

100 per 10s

250 per 10s

500 per 10s

negotiable

Query timeout

30s

30s

60s

120s

negotiable

CORS

restricted

open

open

open

open

Details that matter in production: quotas are counted separately for mainnet versus testnets; queries that exceed the timeout return 504 (so the no-cost tiers effectively cannot run the heaviest queries at all, which Koios documents rather than hides); more than 100 requests per endpoint per 10 seconds from one IP triggers a 60-second 429 sleep; and CORS being restricted on the Public tier means browser-side dApps need at least a registered key.

The payment and legal mechanics are unusual and deserve plain statement. All payments are in ADA, via a CIP-30 wallet-signed flow on the Koios site; there is no fiat option, no invoices (the DAO "is not a legal entity" per its own terms), and no refunds. And there is no SLA at any price point, including Premium: the terms state that Koios does not guarantee any specific service level or uninterrupted operation and encourage users to implement their own redundancy. For a hobbyist, none of this matters. For a company whose accountant needs a VAT invoice or whose customers have uptime obligations, each item is a procurement conversation.

Nexus tiers: Starter $9/month (3M requests/month, 10 req/s with a 100-token burst, 2 keys, no trial), Builder $29/month (10M, 25/250, 4 keys, 7-day trial; every plan is paid), Growth $79 (30M, 50/500, 7 keys), Scale $199 (50M, 100/1,000, 13 keys), Enterprise $499 (200M, 500/5,000, unlimited keys, all add-ons and unlimited WebSocket bundled), Dedicated from $2,000 (physically isolated infrastructure, negotiated contractual SLA, 4-hour support response). Add-ons on every self-serve plan: Market Data $19, Wallet Analytics $15, Transaction Builder $19, MCP $19, IPFS $5, or bundles from $19; WebSocket packages $29/$59/$99. Metering is dual (monthly pool plus per-second token bucket), quota state comes back in X-RateLimit-* headers, add-on gating is a machine-readable 402, and there is no overage billing: the cap is a hard 429 until you upgrade. Billing is conventional (PayPal rails, card-friendly, invoices exist), which is precisely the boring property some finance departments require.

On raw price per on-chain request, Koios wins and it is not close: 500k requests per day for ~$30 versus 10M per month for $29 (or 3M for $9 on Starter). If your workload is chain reads and fits Koios's timeouts and limits, Koios is the rational choice, and we say that as a paying-attention competitor. The Nexus $29 buys a different thing: a smaller pool of requests against a much larger surface, plus a commercial counterparty. The comparison is scope and accountability, not volume.

Reliability

Koios's reliability engineering is better than "community-run" might suggest: health-checked failover across independent instances, tip-synchronization checks, DDoS protection, and a public track record of steady releases (v1.4.0 in March 2026, v1.4.2 in July 2026). But its documented operating envelope is real: 504s on long queries by design, the anti-DoS 429 sleep, 413s on oversized bulk bodies, no public status page we could find, and the no-SLA terms quoted above. The honest summary is best-effort infrastructure run by competent people who explicitly tell you to bring your own redundancy.

Nexus is managed infrastructure with the redundancy built in (the provider chain and tip-staleness routing described earlier) and contracts where contracts belong: Dedicated carries a contractual SLA, isolated infrastructure and a 4-hour response channel. Nexus publishes a public status page with per-service uptime and incident history. It does not yet publish a third-party uptime figure, and we will not invent one here; the 7-day trial plus the rate-limit headers exist so you can measure behavior against your own workload before paying.

When Koios is the better fit

  • Cost-sensitive chain-data workloads. Learning, prototyping, hobby projects, research, indexer backfills: 50k requests/day for free with a key, half a million for ~$30. Unbeatable, full stop.

  • Batch pipelines that fit the PostgREST model. Column selection, rich filters, CSV out, bulk POSTs: for analytical extraction jobs this query model is genuinely more expressive than a fixed REST surface, Nexus's included.

  • Governance tooling. The 15-endpoint Conway group is the broadest free governance API on Cardano.

  • A self-hosting or decentralization mandate. The whole stack is open source and deployable as a bit-identical private instance; there is no vendor to be locked into, structurally.

  • Paying in ADA is a feature for you (a DAO treasury, say) rather than a procurement problem.

When Nexus is the better fit

  • Your product needs anything above on-chain reads. Prices, candles, order books, P&L, transaction construction, swap routing, streaming: none of it exists in Koios, so the real alternative is Koios plus two or three more systems you build and operate. That composite, not Koios itself, is what a Nexus subscription competes with.

  • You have uptime obligations to your own customers. Best-effort upstream means your reliability ceiling is someone else's volunteer capacity, and Koios's terms say so explicitly. Accountability has to flow downhill through contracts.

  • Your queries are heavy or your traffic is bursty. Nexus's per-second buckets with defined bursts, monthly (not daily) pools, and absence of query-timeout tiering behave more predictably under production load than a 100-per-10s IP throttle with 30-to-120-second timeout ceilings.

  • Procurement needs an invoice, a fiat rail, or an SLA. Koios structurally cannot provide any of the three; Nexus provides all of them (the SLA on its Dedicated tier).

  • You want the multi-chain UTXO surface (Bitcoin, Midnight, Apex Fusion, live today) from the same vendor and docs.

Bottom line

Koios is the best value in Cardano chain data, an exemplary open-source project, and structurally incapable of being your accountable vendor; those are all the same fact viewed from three angles. Nexus overlaps it on the chain-data layer, exceeds it everywhere above that layer, and sells the accountability Koios's model deliberately omits. Plenty of serious teams should choose Koios. The moment your product needs market data, analytics, transaction construction, streaming, or a contract behind your data layer, the comparison stops being close.

Evaluate the full surface against your own traffic with the 7-day Builder trial: get a key at nexus.gerowallet.io, browse the live endpoint playground at nexus.gerowallet.io/docs, or contact us about Enterprise and Dedicated via the support page.

Still comparing? Read Nexus vs Blockfrost for the other half of the shortlist.