Documentation based guide: compiled from each maker's own docs and pricing pages on Oct 30, 2026. We have not load-tested these endpoints, and the guide contains no code we ran.
For most TON projects, start with TON Center API: it is the API the TON docs call the official HTTP provider, it serves mainnet and testnet, and a free key from its Telegram bot gives 10 requests per second. Use TonAPI when you want decoded Jetton, NFT and wallet data without parsing cells yourself, and move to a managed provider (Chainstack, QuickNode, GetBlock) when you need higher throughput, archive history on one bill, or a dashboard your team already uses.
This guide goes deeper than our TON infrastructure page: which TON API versions each option serves, published free limits, and what changes when you need full history. It summarizes what each maker says; it is not a ranking.
Contents
- Quick answer
- What do "API v2, v3 and v4" mean on TON?
- TON Center API
- TonAPI (TON Console)
- Chainstack
- QuickNode
- GetBlock and NOWNodes
- Running your own liteserver
- How to choose
Quick answer
| Option | TON APIs served | Networks named | Free access | Source |
|---|---|---|---|---|
| TON Center API | v2 (liteserver), v3 (indexer), Streaming v2, Emulation | Mainnet, Testnet | 1 req/s without a key; 10 req/s with a free key | Source: TON docs |
| TonAPI | REST with decoded data, Webhooks (Streaming API deprecated), liteserver access | Mainnet, Testnet | 1 request per 4 seconds without a token | Source: TON Console docs |
| Chainstack | v2, v3, v4 | Mainnet, Testnet | Developer plan: 3M requests/mo, 25 req/s | Source: Chainstack pricing |
| QuickNode | HTTP API v2 and v3, archive | Mainnet only | 1-month trial, 10M credits, 15 req/s | Source: QuickNode TON docs |
| GetBlock | JSON-RPC, HTTP API v4, Indexer API v3 | Mainnet (free); Mainnet and Testnet on dedicated | 50K compute units a day, shared mainnet endpoint | Source: GetBlock TON page |
| NOWNodes | RPC, WebSocket and archive on supported networks | Mainnet on its TON page | Start plan: 100,000 requests/mo, 1 key | Source: NOWNodes pricing |
| Your own liteserver | Raw liteserver (ADNL) access, plus self-hosted v2/v3 | Mainnet, Testnet | Your hardware | Source: TON docs |
What do "API v2, v3 and v4" mean on TON?
TON has no single JSON-RPC standard like Ethereum's, so "TON RPC" can mean different things. The TON docs list three ways to read data: public liteservers, TON Center API v2 and TON Center API v3. In their comparison, v2 sits on top of a liteserver, while v3 is an indexer that "maintains its own database derived from blockchain data for richer queries (traces, jettons, NFTs, etc.)". Liteservers are the only option of the three that return cryptographic proofs (Source: TON docs API overview).
Chainstack's docs add a v4: block-pinned reads where each account-state or get-method call names a masterchain block seqno, which is what `TonClient4` in @ton/ton uses. Their rule of thumb: v2 for the freshest data, v3 when you can accept a small indexing delay for cleaner data, v4 if your code uses TonClient4 (Source: Chainstack, choosing v2, v3 or v4).
Gotcha: a provider that only serves v2 and v3 won't work with code written for TonClient4.
TON Center API
The TON docs describe TON Center as "the official provider of HTTP APIs for TON". It serves API v2 at toncenter.com/api/v2 and v3 at toncenter.com/api/v3, with testnet.toncenter.com for Testnet, plus a Streaming API over SSE and WebSocket and an Emulation API for simulating transactions before you send them (Source: TON docs API overview).
Limits. Without a key you get 1 request per second. The plans are Free (10 req/s, 1 key per network), Plus (25 req/s, 3 keys), Advanced (100 req/s, 10 keys) and Enterprise (custom). Limits apply to all your keys combined, separately for mainnet and testnet, and over-limit calls return a 429 (Source: TON Center rate limits).
Trade-offs. Keys and upgrades are handled only through the @toncenter Telegram bot, and paid plans are priced in GRAM (Plus 2.5 GRAM/month, Advanced 25 GRAM/month) and paid by sending a transaction (Source: TON docs, get an API key). Requests without a valid key fall back to 1 req/s even on a paid plan. History depth on v2 "depends on liteserver"; v3 is listed as archival.
TonAPI (TON Console)
TonAPI, from the Tonkeeper team, is built to hide TON's low-level details: its docs say you don't need to worry about "supporting all wallet standards, performing verifications, or decoding raw bits and bytes". The docs cover a REST API, a Webhooks API, an SDK and paid liteserver access (Source: TON Console docs). Its Streaming API page now says it is deprecated and points to the Webhooks API for real-time updates; Webhooks need a private API key (Source: TON Console Streaming API, Webhooks API).
Limits. All methods work without a token at 1 request per 4 seconds (0.24 rps). Tokens come from tonconsole.com, and your plan's rate applies on mainnet and testnet alike. The plan prices live on TON Console's pricing page, which we couldn't read as text, so we've left them out.
Self-hosting. opentonapi is an open-source, limited version. The docs warn that some data needing a full index, "such as (jettons, NFTs, and full account information)", may not be visible in it.
When it fits: wallet UIs, NFT and Jetton views and payment flows. For verified raw data, its liteserver product gives ADNL access to two liteserver instances with historical data (Source: TON Console liteservers).
Chainstack
Chainstack lists TON on Global Nodes and Dedicated Nodes, with archive data on both Mainnet and Testnet Global Nodes (Source: Chainstack protocols and networks). It serves TON API v2, v3 and v4 at the same price. Most calls cost 1 request unit; calls that read history (getTransactions, and block reads 128 or more blocks behind the tip) count as archive and cost 2 units (Source: Chainstack TON docs). Native ADNL access is a Dedicated Node feature.
Limits. The free Developer plan includes 3M requests a month at 25 req/s; Growth is $49/month and the pricing page lists "Archive Data" from Growth up (Source: Chainstack pricing). Check how that applies to TON history calls before relying on the free plan. Compare it with GetBlock in Chainstack vs GetBlock, or with the official API in TON Center API vs Chainstack.
QuickNode
QuickNode serves TON Mainnet only. Its TON docs list Mainnet with archive "Yes" and pruning "None", and an API reference for the TON HTTP API v2 and v3 plus a validators rewards API; the docs carry a "Beta" label (Source: QuickNode TON docs). Its TON page also mentions Streams for TON (Source: QuickNode TON page).
Limits. A 1-month free trial with 10M API credits at 15 req/s, no card; then Build at $49/month ($34/month billed yearly) (Source: QuickNode pricing). For testnet you will need another endpoint, such as TON Center or Chainstack. See also QuickNode vs Chainstack.
GetBlock and NOWNodes
GetBlock's TON page offers JSON-RPC, HTTP API v4 and Indexer API v3. The free plan is a shared TON mainnet endpoint with 50K compute units a day and no card; shared plans start at $49/month (250 to 800 req/s tiers); dedicated TON nodes, with Mainnet and Testnet, start at $1,700/month for a full node and $2,500 for archive (Source: GetBlock TON page).
NOWNodes lists the network under the name Gram on its TON node page, with Mainnet shown and RPC, WebSocket and archive described for "supported networks" (Source: NOWNodes TON page). The free Start plan gives 100,000 requests a month and 1 API key (Source: NOWNodes pricing). Its TON docs page didn't state which TON API versions it serves, so test your SDK calls on the free plan first. The two are compared in GetBlock vs NOWNodes.
Running your own liteserver
Running your own node gives you proofs, no shared rate limits and control over history. The TON docs set a high bar: a liteserver needs a 16-core CPU, 64 GB RAM, at least 1 TB of NVMe Gen4+ storage with 64,000 provisioned IOPS, 1 Gbit/s symmetric bandwidth (about 16 TB a month at peak) and a static IP (Source: TON docs, run a liteserver). An archive liteserver needs 128 GB RAM and at least 16 TB of storage with ZFS lz4 compression, or 20 TB without (Source: TON docs, run an archive liteserver).
On top of your liteserver you can deploy TON Center's open-source v2 and v3 services, so your code doesn't change when you move off the hosted API.
How to choose
- Prototype or Mini App MVP: TON Center with a free key (10 req/s, mainnet and testnet). Add a backoff on 429s.
- Wallet, NFT or Jetton UI: TonAPI's decoded REST data, or TON Center v3 if you prefer the official indexer.
- Your code uses TonClient4: pick a provider that serves v4 (Chainstack, GetBlock).
- Full history in production: QuickNode (mainnet archive, no pruning), Chainstack archive (2 units per historical call) or GetBlock dedicated archive; check what each counts as an archive call.
- Testnet as well as mainnet: TON Center, TonAPI or Chainstack; QuickNode lists Mainnet only, GetBlock's free plan is mainnet.
- Fiat billing and a team dashboard: Chainstack, QuickNode, GetBlock or NOWNodes rather than TON Center's Telegram-bot plans.
- Proofs or no shared limits: your own liteserver, or a paid liteserver plan.
For every TON tool we list, with the page each maker's docs name the chain on, see TON developer tools.