The fastest, most reliable path to the Base sequencer.
Every transaction submitted to Azul — free tier or paid — fans out to a co-located local node and the fastest healthy upstreams in parallel. Whichever path reaches the sequencer first wins. No relay, no detour, no queue.
Free
The default tier. Same parallel submission path, same co-located node, same routing as Edge. No key, no signup, no billing surface.
No card, no email. Paste the URL and go.
Use the public endpoint- Parallel race-to-land submissionTx mirrored to a co-located local node + top upstreams concurrently — first path to include wins
- Sandwich-resistant by defaultRace-to-include minimizes the pending window an attacker can front-run
- Co-located local nodeSubmissions land on a tuned reth peered with the proxy on the same box
- Latency-weighted read poolTop-3 of 16 Base RPCs, re-ranked every 15s
- Public RPC URLhttps://rpc.baseazul.dev
- Rate limit100 req/s per IP · burst 400
- Heavy methods10 req/s per IP (eth_getLogs, large eth_call)
- Batch size50 requests per call
- eth_getLogs5,000 blocks per call
- MethodsAll standard eth_*, net_*, web3_*
- LoggingNone — no IPs, no bodies, no headers, no cookies
- CORSAllow-Origin: * — works from any origin
- SupportCommunity
Trading desk
Everything the free tier gets, plus the data and access that pay for themselves on the first fill: pre-confirmation deltas, live mempool, hot state, privileged methods.
By invitation. Limited seats to keep contention low on the upstream pool.
Request access- Private submission laneOptional — write path bypasses public upstreams. Direct sequencer reach, zero mempool footprint.
- Pre-confirmation Flashblocks feedBlock deltas streamed before finalization
- Live mempool & log streamPending-tx and log firehose over a private WS
- Priority submission laneYour tx jumps the queue on every parallel path
- Direct hot-state read peeringYour trading client peers with the local node — sub-ms hot reads
- Privileged namespacesdebug_*, trace_*, txpool_* unlocked
- Rate limitUnlimited — your key bypasses every cap
- Batch size500 requests per batched call
- eth_getLogs100,000 blocks per call (20× public)
- Max response500 MB (10× public)
- Daily egressUncapped
- Per-key analyticsLatency, hit-rate, propagation timing exports
- IP allowlistPin the key to your fleet
- OnboardingArchitecture walk-through with the operator
- SupportDirect Telegram / Signal · 15-min ack
The engineering doesn’t care which door you came in.
The thing that makes Azulthe fastest RPC to the Base sequencer isn’t the tier — it’s how the proxy is built. Free users hit the same routing, the same write path, the same tuning as Edge.
Parallel write fan-out
eth_sendRawTransaction is mirrored to the local node and the fastest healthy upstreams at the same instant. Whichever path lands first wins — fewer dropped txs, lower inclusion tail-latency. Free and Edge use the same fan-out.
Sandwich-resistant submission
Sandwich attacks need a long pending window to wedge in front of your tx. Race-to-include kills that window for free users. Edge customers can opt into a private lane that skips public upstreams entirely — zero mempool footprint.
Co-located local node
The proxy runs on the same machine as a tuned reth instance. Your raw tx clears userspace once, hits the local node directly, and mirrors out — no internal hop, no extra TLS, no relay.
Latency-weighted routing
Reads load-balance across 16 Base RPCs. A 15-second health probe re-ranks them; queries hit a random selection from the top-3 fastest. Members that slow down silently drop out.
Pre-warmed connections
Persistent HTTPS keep-alive to every upstream. No cold TLS handshakes during a burst — the proxy holds the sockets so your first request lands as fast as your hundredth.
Streaming response path
Compression (zstd, gzip) negotiated per-client. Responses pipeline end-to-end without buffering the full body — wire bytes start moving before the upstream has finished writing.
Single region, no SLA
Everything above runs from one location (Virginia) today. Edge customers get priority routing and direct support, but no formal uptime contract. If your workload needs multi-region or a signed SLA, mention it when you request access.
What we don't publish
A few specific propagation tricks live in the source and stay there. We tell you what they do (faster, more resilient submission). We don't write a how-to that helps a copycat reach parity.
Most people stay on Free.
Edge is for desks running fill-sensitive strategies — pre-confirmation data, mempool, debug methods. The public tier doesn’t expose any of that, and most wallets / dApps don’t need it. They get the same fast tx path either way.