Our node infrastructure
Every response from the 1st Node API is served by hardware we provision and operate ourselves — no reseller in the path, no hidden shared rate limit. Here is exactly what runs the fleet and the targets we hold it to.
In-region, cached-hot reads (eth_getBalance, eth_call).
Cold archive reads under load, single region.
Rolling 30-day per-chain endpoint availability.
Historical state from block 0 on archive-tier chains.
Anycast routing to the nearest healthy replica.
Every chain runs at least two independent nodes.
Fleet & archive coverage
Archive nodes are expensive to run, so coverage is where providers quietly cut corners. We label every chain honestly: archive means full historical state from genesis, full means a complete non-pruned node without deep trace.
| Chain | Client | Tier | Trace | Detail |
|---|---|---|---|---|
| EthereumETH | Erigon + Lighthouse | archive | Full archive from genesis; trace_* and debug_* served on the same endpoint. | |
| ArbitrumETH | Nitro (archive) | archive | Archive Nitro node with classic pre-Nitro state; arbtrace_* available. | |
| OptimismETH | op-geth + op-node | archive | Archive op-geth with Bedrock history; debug_traceTransaction served. | |
| BaseETH | op-geth + op-node | archive | Same Bedrock stack as Optimism, archive-mode, full trace support. | |
| PolygonPOL | Bor + Heimdall | archive | Archive Bor node; trace_block and debug_* over full history. | |
| BNB Smart ChainBNB | bsc-geth | archive | Archive bsc-geth; large state footprint served from NVMe. | |
| AvalancheAVAX | AvalancheGo (C-Chain) | archive | C-Chain archive with full historical state; debug tracing enabled. | |
| BitcoinBTC | Bitcoin Core + txindex | archive | Full node with txindex=1 for any-txid lookup; no pruning. | |
| TRONTRX | java-tron (full) | full | Full node with event/transaction history; deep-archive queries via internal index. | |
| TONTON | ton-http-api + validator | archive | Archive liteserver serving historical masterchain and shard blocks. |
How we run it
Our own metal, not a reseller
Every node in the fleet is bare-metal hardware we provision and operate. We do not proxy a managed RPC provider, so there is no third party between your request and the chain — and no shared rate limit you cannot see.
Honest archive, honestly labelled
Archive nodes are expensive, so many providers quietly prune history and 404 on old state. We run full archive where we say we do and label anything that is a full-node-only tier, so trace and historical-state queries either work or are documented as unavailable.
Redundant by default
Each chain runs behind at least two independent nodes with health-checked failover. A node resync or host failure drains traffic to the healthy replica without a client-visible outage.
Measured, not asserted
Latency and availability are sampled continuously from external probes, not read off a status page. The numbers above are the targets we hold the fleet to and alert on when breached.
How we measure
Latency percentiles are sampled continuously by external probes issuing representative reads against each regional endpoint; availability is the rolling 30-day fraction of successful health checks per chain. The figures above are the SLO targets we alert on when breached — not a one-time best-case screenshot.
Top up, get a key, ship.
Self-serve. Pay in crypto or card. Metered by credits — heavy primitives cost more, simple ones are cheap.
Get API key