Key figures from our checks
The table below compares three TON API endpoints on uptime, response time and block lag, taken from our own recurring checks.
Each row shows how many checks sit behind its percentages, and that count matters. A handful of samples can swing uptime by several points, so read the percentage and the check count together. How the checks work is explained further down, and the full history lives on the Pulse page.
| Target | Uptime, 7 days | Uptime, 30 days | Median response (p50), 7 days | Slow tail (p95), 7 days | Median block lag, 7 days | Checks, 7 / 30 days | Last check |
|---|---|---|---|---|---|---|---|
| TON Center API v2 | 100% | 98.33% | 182 ms | 523 ms | 4 blocks | 32 / 60 | Up |
| TON Center API v3 | 100% | 98.33% | 149 ms | 512 ms | 0 blocks | 32 / 60 | Up |
| TonAPI | 100% | 100% | 129 ms | 511 ms | 3 blocks | 32 / 60 | Up |
Values refresh whenever the data file does. The three probes call different endpoints, so a gap in response time between rows is a hint, not a ranking.
Data as of · Source: Pulse bot, checks several times a day from a GitHub-hosted runner in the US, check counts in the table · How we measure
Toncenter vs TonAPI: what is the difference?
TON Center is the provider documented on docs.ton.org. TonAPI is a separate REST service with its own keys.
docs.ton.org calls TON Center "the official provider of HTTP APIs for TON" and lists three products: API v2, API v3 and a Streaming API v2. TonAPI describes itself as a SaaS service with REST, streaming, webhooks, GraphQL and liteservers. Its footer still reads "Powered by Tonkeeper"; that wallet is now Keeper (formerly Tonkeeper).
Inside TON Center the split is about indexing. API v2 is the non-indexed layer: it uses tonlib to query liteservers and speaks REST and JSON-RPC. API v3 reads raw node storage, parses it and keeps it in PostgreSQL, so it suits history queries better than v2. On v2, minor version 2.0.x is the older Python build and 2.1.x is the C++ build the docs call newer and recommended.
| Item | TON Center API v2 | TON Center API v3 | TonAPI |
|---|---|---|---|
| Layer | Non-indexed, relays to liteservers | Indexed, data stored in PostgreSQL | Hosted REST service |
| Interfaces | REST and JSON-RPC | REST, grouped by accounts, traces, jettons, NFTs, DNS, staking and more | REST, webhooks, streaming, GraphQL |
| Documented use | Balances, state, get-methods, sending messages | Transaction history, traces, decoded Jetton and NFT data, analytics | Accounts, jettons, NFTs, emulation, gasless |
| Indexer and archive (per docs.ton.org) | No indexer; archive depends on the liteserver | Indexer and archive | Not covered by docs.ton.org |
| Self-hosting | Open source, can be self-hosted | Open source, can be self-hosted | Limited open-source version (opentonapi); some data may be missing |
| Base URL or reference | toncenter.com/api/v2 | toncenter.com/api/v3 | Swagger UI at tonapi.io/api-v2 |
| Streaming | Separate Streaming API v2 (SSE and WebSocket, API key) | Same Streaming API v2 | Streaming API marked deprecated in TonAPI docs; Webhooks API for real-time updates |
One more difference sits in real-time data, and the sources disagree about it. TON Center documents Streaming API v2 over SSE and WebSocket. TonAPI's own docs mark its Streaming API as deprecated and point to its Webhooks API, while docs.ton.org still lists TonAPI streaming endpoints. Each source says its own thing, so check the page for the endpoint you plan to use.
How we measure speed and uptime
We send one light request to each TON API from a US server several times a day and record status, time and block number.
The runner is a GitHub-hosted machine in the US. A check times out after 10 seconds and counts as slow past 3 seconds. The three probes are a getMasterchainInfo call on API v2, a masterchainInfo call on API v3 and a masterchain-head call on TonAPI. Rate-limited and blocked answers are tracked apart from plain outages in our data file.
Those three calls do different work. v2 relays to a liteserver, v3 reads from its indexer database and TonAPI serves its own REST layer, so one sample is not like for like across them. TON Center states on toncenter.com a 99.98% mainnet availability over 90 days and an 85 ms median response. Its vantage point differs from our single US runner, so the two sets are not comparable. Method details are on the methodology page.
Do I need a TON API key?
You can call both without a key, but a TON API key raises the limit. TON Center allows 1 request per second without one.
A key is free on TON Center. Open the @toncenter bot in Telegram, press Start, choose Manage API Keys, pick a plan (Free is the default) and create the key. Mainnet and testnet get separate keys. On API v3 the key goes in the X-API-Key header or the api_key query parameter, and the docs say never to expose it in client-side code. Prices are quoted in Gram (GRAM, also known as Toncoin).
| Plan | Requests per second | Keys per network | Price | Notes |
|---|---|---|---|---|
| No key | 1 | none | free | Default limit |
| Free | 10 | 1 | free | Shared liteservers, low-volume testing |
| Plus | 25 | 3 | 2.5 GRAM per month | Private liteservers |
| Advanced | 100 | 10 | 25 GRAM per month | Private infrastructure |
| Enterprise | custom | custom | contact @toncenter_support | Custom throughput and support |
Limits count all keys of one account together, separately for mainnet and testnet, so three Plus keys on mainnet share 25 requests per second. Above the limit, TON Center answers HTTP 429 with the body {"ok": false, "result": "Ratelimit exceed", "code": 429}. The docs advise stopping, waiting a few seconds and retrying with exponential backoff.
TonAPI keys are issued at tonconsole.com and sent as an Authorization: Bearer header. For access without a key, the docs disagree with each other. The TonAPI intro gives 1 request per 4 seconds, while the free limits page gives a default of 1 request per second. We do not pick one. Limits are the same for all tokens on a plan and for testnet and mainnet. TonAPI also lists a separate price list for Liteservers on tonapi.io.
| Plan | Requests per second | Price listed |
|---|---|---|
| Start | 1 | Free |
| Lite | 10 | $9.9 |
| Standart | 100 | $95 |
| Plus | 300 | $290 |
| Heavy | 1,000 | $890 |
| Custom | unlimited | contact support |
Wallet, bot or Mini App: which API?
Any TON API can serve a wallet, a bot or a Mini App. The choice comes down to which data your app reads most often.
Choosing a TON API starts with the call you make most. TON Center documents v2 for balances, state and sending messages, and v3 for history and decoded tokens. TonAPI lists accounts, jettons, NFTs, emulation and gasless transfers on its homepage, with keys managed in TON Console. Hosted node access is a separate product, for example dTON with private liteservers.
- Wallet: balances and sending are v2 territory in the docs; emulation, which TON Center and TonAPI both list, lets you preview a transaction before it goes out.
- Bot: in our reading of the docs, a bot that reacts to events needs a push channel. TON Center has Streaming API v2, TonAPI has Webhooks.
- Mini App: the page runs in a user's browser, so a key placed there is visible. TON Center's docs say to keep keys out of client-side code. TonAPI documents a route through the Keeper browser or extension with higher limits, which still needs a valid key. Telegram's own docs are on the Mini Apps card.
Block lag: why fresh data differs
Block lag is how many masterchain blocks an API sits behind the freshest of the three in the same check.
Every API reads the newest block from a different path: v2 asks a liteserver through tonlib, v3 serves what its indexer has already parsed and stored, and TonAPI answers from its own service. Each extra step between node and response is a place where the newest block may not have arrived yet. See the glossary for Indexer and Liteserver.
We have not measured which step adds the delay, so we do not claim a cause. The Key figures table above lists each API's median lag against the freshest API over the window. A gap of a few blocks can come from requests landing at slightly different moments, so a single check cannot rank services. If your app compares block numbers or waits for a transaction to appear, a retry is safer than trusting one read.
How to check it yourself
You can repeat our probes with curl in a minute and compare the time and block number each API returns.
Use the same three TON API endpoints we do. Without keys you are inside the no-key limits above, so a few seconds between calls are needed: TON Center answers HTTP 429 if you go faster. Your location changes response time, so your numbers will not match ours, and that is the point of running them. More tools around these services are in the developer tools category of our directory of TON projects.
- Time TON Center v2:
curl -s -o /dev/null -w "%{time_total}\n" https://toncenter.com/api/v2/getMasterchainInfo - Time TON Center v3 the same way against https://toncenter.com/api/v3/masterchainInfo
- Time TonAPI against https://tonapi.io/v2/blockchain/masterchain-head
- Drop the timing flags and read the block number in each JSON answer. Run the three calls a few seconds apart, then repeat later in the day.
- Compare your times and block-number gaps with the table above, and open our raw Pulse data if you want the individual checks.
FAQ
Is TON Center free?
Yes, within limits: 1 request per second without a key, 10 with a free key, per docs.ton.org.
What is TON Center API v3?
The indexed layer: it stores parsed node data in PostgreSQL and serves history, traces and decoded Jetton and NFT data.
Does TonAPI need an API key?
Not strictly. Without one, its docs give two limits: 1 request per 4 seconds, and elsewhere 1 per second.
How often do you check?
Several times a day from a US server. Each row in our table shows the number of checks behind it.
What does a 429 error mean here?
You went over your limit. TON Center advises stopping, waiting a few seconds and retrying with exponential backoff.
Projects in this guide
Every project named in this guide has its own page with a short description, live data where we collect it and a link to its official site.




