Nodely's free Algorand API answers 115M+ requests a day — slowed by design

2026-08-20

infrastructure · api · indexer · node · discovery · web · algorand · updated

Nodely's free Algorand API answers 115M+ requests a day — slowed by design

The shared window onto the chain

Building anything on a blockchain begins with a decision that has nothing to do with the product itself: who reads the chain on your behalf. Every balance check, transaction lookup and smart-contract read has to reach a node, and running your own means paying to keep a synced copy available around the clock. Nodely, the Warsaw company most Algorand developers still know by its former name AlgoNode, turned that chore into a business and gives the most common version away. Its homepage puts the free tier's scale at 115M+ API requests per day, served across 75+ indexers in more than 20 geographic locations, with 75+ customers on five-plus AVM-compatible chains.

The surrounding evidence broadly supports the picture. The free mainnet endpoint answered a live status request just a few rounds behind the network head during reporting, the indexer returned current transactions, and the public status page showed the free node API at 99.999% uptime and the free indexer at 100%, with no incidents logged in the week before this writing. A live 3D globe renders the physical footprint — nodes reporting from Virginia, Oregon, Mumbai, Singapore, Johannesburg, Santiago and central Europe. The one soft spot in the record is the light-indexer snapshot service, at 92.952% — a reminder that '5 bars - always' is a slogan, not a guarantee.

A self-funded team, a renamed company

The operator is Nodely Sp. z o.o., incorporated in Warsaw at Żwanowiecka 53 and listed in the Polish National Court Register (KRS 0001061171). Its FAQ describes the company in two sentences: 'We are a self-funded team with the aim of providing the best native, globally distributed, low latency Algorand API service.' And, asked about investors: 'We have no investors. The team pays all the bills.'

The AlgoNode rebrand was, by the company's own account, cosmetic — the FAQ answers the difference question with 'There's no difference!', and the legacy algonode.cloud endpoints keep running under the same rules and infrastructure as the newer nodely.dev URLs. The engineering footprint is visible on the AlgoNode GitHub: 37 public repositories, and the team still pushes to the core ones — the block-streaming server, the light indexer, the CockroachDB conduit plugin — as recently as August 17, 2026.

What the free tier actually is

Getting started requires no API key and no registration: a developer drops one of Nodely's endpoint URLs into the standard JavaScript or Python SDK with an empty token field and starts reading chain state immediately. The free endpoints cover Algorand mainnet, testnet, betanet and FNet, plus the Voi network, and the docs note the old algonode.cloud URLs remain valid alternatives. An IPFS gateway for Algorand asset metadata rides along on the same tier. Nodely describes the free tier as production-usable, asking only that projects doing more than 1M requests a month credit it on their site.

What the docs call the fair-use policy is where the free lunch gets its portions. The limits apply per IP address and per project, and Nodely reserves the right to reject requests — returning HTTP 429 for rate limits and 403 for daily quota breaches — once they are exceeded:

Concept	Real-World Implication
60 requests/s per IP, 10 concurrent calls	A browser or lightly used backend stays inside the envelope; anything heavier needs caching
5 transaction submissions/s per IP	Write flows are throttled harder than reads — deliberate, since each submission is a broadcast to the network
5 GB of responses per day, 150 GB per month	An indexer-heavy app burns through bytes long before it touches the request limit
6M requests/month and 1,000 req/s per site	The ceiling for a popular front-end before a paid tier becomes relevant

The policy's bluntest advice is to cache immutable history: the FAQ warns that repeated identical requests with min-round=0 or 1 will get a backend blacklisted. That guidance is worth reading as a design philosophy — the free tier only works if developers stop re-fetching data that cannot change.

The free tier's real price

'Free' here is a throttle, not just a price tag. The quickstart page states that free-tier endpoints inject artificial latency into every response, while the fair-use policy puts the delay at 150ms (250ms for archival) — the two pages do not agree on the number, which says something about doc hygiene even though the mechanism is unambiguous: the free tier is deliberately slower, and commercial endpoints remove the delay entirely.

There is a second, less advertised cost. The free tier's per-project usage is published on a public Grafana dashboard, and the FAQ itself lists 'they would like to hide the API call logs from the public dashboard' among the reasons teams buy the paid tier. Whatever traffic a free-tier app generates is, in effect, public telemetry; a team that does not want its usage patterns visible has to pay.

The same dashboard makes the ecosystem's dependency concrete: in the trailing 30 days it shows randgallery at 140% of its monthly bytes quota, metrics-iframes at 69% of its request quota, and statto, exa, chaincontrol, algofaucet and a dozen other named projects consuming meaningful fractions. As this publication covered in August, the explorer ecosystem — Allo.info, Lora and Pera among them — sits on the same free infrastructure. One private company is, in effect, the default reader for a large share of the network's public traffic.

What pays for the giveaway

Nodely's stated model is a ladder, not a wall. 'The primary goal of Nodely is to help teams run their own API infrastructure,' the FAQ says, and teams move up when they need a guarantee, a phone number, dedicated resources — 'But mostly because Nodely has helped them become profitable and now they can afford it.' The paid menu is fixed-price, with no overages:

Concept	Real-World Implication
Unlimited — $256/month	Shared-infrastructure access at up to 6,000 req/s per key across 25 regions, a 2 TB monthly quota and a 99.99% SLO; billed by card, SWIFT or USDCa
Business — $3,072/year	A 100% SLA with tenfold reimbursement, four-hour response, fee abstraction at $0.001 per transaction with 1M transactions included, and a validator node service
Enterprise — $1,024/region/month	Dedicated bare metal per region (NA/EMEA/APAC/LATAM), vanity domains, 180-day log analytics, 10x reimbursement across three regions
Analytics DB — $2,048/month	Shared ClickHouse cluster with Superset self-service BI and queries against any historical block height
App Chain — $4,096/month/network	A dedicated hybrid Algorand network — hosted, managed, public or private — with a dedicated SecOps team
SQL Direct — custom quote	Dedicated PostgreSQL per region (32 vCPU, 128 GB RAM, 8 TB NVMe) with an Algorand indexer schema, real-time NFT metadata parsing for ARC-3/19/69 and ~100 ms block indexing

The no-lock-in promise runs through all of it: the APIs are vanilla Algorand node and indexer interfaces, the patches are public, and the open-source light indexer can stand up a full indexer on a 25 GB SSD in about 20 minutes — the escape hatch that makes 'migrate to your own setup' a concrete offer rather than a slogan.

Counting the network the network can't count

Nodely's most influential product may not be an API at all — it is a number. Since 2024, ephemeral home nodes and short-lived bot connections made the old unique-IP node count meaningless, so the team adopted the Chao-1 estimator, a statistical technique from ecology for estimating population size from incomplete samples, and applied it to hourly observations of which IP addresses connect to its relays. The result is a daily estimate of full-time node equivalents — and it has become official. The Algorand Foundation's node-count KPI, co-written by Nodely founder Paweł Pierścionek and the Foundation's head of business intelligence, states flatly that 'The Nodely analytics dataset is the source of the node count KPI,' exposed through a public API. That API returned 2,695 node equivalents for August 19, 2026, drifting up gently through the preceding fortnight:

Chart: Full-time Algorand node equivalents, Aug 6–19 2026

The methodology page is candid about the method's limits — an IP address is not a node (NAT and CGNAT hide many), dynamic IPs and Sybil farms skew samples, and nodes on private relays are invisible. What the page does not dwell on is the structural tension: the same company operates relays, runs the telemetry service that feeds the leaderboards, sells enterprise access, and produces the official network-size figure. The KPI is Nodely's measurement of a network that includes Nodely's own infrastructure — reasonable work, but an estimate from an interested party.

Data, staking and the lines in between

The API business shades into public goods and into services with higher stakes. Since June 2024 the company has published its commercial BigQuery dataset of the full mainnet — transactions, accounts, blocks, token prices and holder tables — free to query. Daily archival snapshots of the whole chain sit in a public S3-compatible bucket at up to 10 Gbps download, free even for service providers (kept three days, with an honest warning to treat any single source of history with care). Its telemetry service gives node operators health scores, voting-performance and synchronization metrics tied to anonymized GUIDs, ingests from 20 data centers, and collects no personal data. The Allo.info explorer, Nodely's analytics platform, layers address-monitoring alerts and plain-language transaction explanations on top of the same data.

The stakes rise with staking. reti.nodely.io hosts a build of the open-source Réti Staking Dashboard, showing live pools with stakes in the hundreds of thousands of ALGO; Nodely's disclaimer frames it as a technology provider only — it holds no crypto, gives no investment advice, and notes the pool contracts have been commercially audited. Then there is AVAIL, the Algorand Validator Adaptive Infrastructure Layer launched in January 2025, which runs and rotates customers' participation keys for $32/month per key at the eight-key tier (or $64 for a single key). Participation keys are the consensus credentials that let an account propose and vote on blocks; they cannot spend the account's funds. That makes AVAIL non-custodial in the strict sense — 'Nodely has no technical means to spend or manage customer's funds,' per its FAQ — but the trust transfer is real: the customer's rewards eligibility now depends on a third party's key security and rotation discipline. Nodely sells that trust on a 60-second-per-month unavailability SLA, and only to business entities that accept invoices.

The concentration question

Nodely is a rare infrastructure story in crypto: a genuinely useful, genuinely free service that the ecosystem has built on top of, funded by the same team that runs it, with no investors to answer to and no token to promote. The concentration is real, and it is the story's honest risk — one Polish company is the default reader for much of the network's public traffic, the keeper of the official node-count figure, and the operator of key material for paying validators. What mitigates it is the company's own design: vanilla APIs, published patches, an open-source indexer anyone can run in twenty minutes. As long as that migration path stays real, Nodely's central position is a convenience the ecosystem chose — not a lock-in it cannot escape.

Source

Source: https://nodely.io/