Network

Built to keep running, and open to anyone who wants to help

A receipt in milliseconds is only worth something if the network behind it never loses a transaction. Here is how it does that, and how you can take part.

Four kinds of node share the work. Each has one job, and losing any of them costs you time, never money.

Contact the team Read the operator guide

An agent paying for inference at two in the morning does not get to retry tomorrow. It needs an answer now, and that answer has to still be true in an hour. So every transaction is written down in two places before anyone is told it happened, checked by independent validators before it is final, and served from enough public endpoints that no single machine sits in your critical path.

public RPC nodes you can read from and send through

< 1 s

from sending a transaction to final, with no reorganization

URL for the whole network, if you only want one

Four roles, one job

The sequencer

Executes each transaction as it arrives and hands back a signed receipt, then proposes blocks in exactly that order, so what you were told is what gets committed.

The hot standby

Holds a byte-for-byte copy of every transaction before your receipt is released. If the primary fails, it takes over in about a second with nothing missing.

Validators

Independently re-execute and commit every block, refusing any that deviates from the receipted order. That is what makes a receipt a promise rather than a hope.

Archive nodes

Keep the full history and serve proofs about any past balance or event, so an auditor can check last year's payment as confidently as today's.

What that means when something breaks

Suppose the sequencer loses power mid-afternoon. Every receipt it issued is already on the standby, so no user sees a payment vanish. The standby takes over and your code keeps sending to the same URL. If the standby is unreachable, the network stops issuing receipts rather than issue ones it cannot keep. You get a delay, never a false success.

Validators are the second line. Because they refuse to commit anything that disagrees with receipts already given out, history cannot be quietly rewritten, and anyone holding a receipt can prove it.

Running a node yourself

01
Install with one command
A single installer fetches verified builds for x86-64 or Arm and the live chain configuration. Downloads are checked against published checksums first.
02
Choose full node or validator
A full node serves reads and relays transactions for your apps. A validator also stakes PAX, takes part in commits, and earns a share of fees and the scheduled PAX release.
03
Sync fast
State sync brings a new node to the current height from a snapshot, so a fresh machine is useful the same day.
04
Stay current
Upgrades are announced on chain at a named block height. One command fetches the build; the node switches over at that height.