gossip topologyslot reserved
Your Hyperliquid nodehl-visor, non-validator, Mainnet4001 / 4002spectrum-peer-01primary
spectrum-peer-02fail over
reserved_peer_ipsyour node IP, allowlisted on both
Hyperliquid peeringDedicated gossip peers for the Hyperliquid node you run yourself
You keep operating your own node. Spectrum reserves a peer slot for it on two of its Hyperliquid peers, so your node has a stable upstream instead of competing for public root peers. 99.9% uptime, flat $400 per month, per node IP.
Your Hyperliquid nodehl-visor, non-validator, Mainnet4001 / 4002$400
Per month, per node IP
Flat rate. No bandwidth or egress charges
2 peers
A held slot on each
Point at either one, fail over to the other
Your node keeps a slot we hold open
Peering is a single, flat subscription. You send us the public IP of the node you already run, we reserve it on two of our Hyperliquid peers, and you point your visor at them.
Single region
Both peers run in one region, so you get redundancy across hosts rather than across datacenters. If you need a peer in a second geography, talk to us before you subscribe and we will confirm what is available.
1 allowlisted node IPYou send the public IP of your node. Spectrum reserves it on both peers, so a slot is held for it whether or not the network is busy.
2 Spectrum peersPoint your node at either one. The second is there so you can fail over without going back to the public peer list.
Full gossipThe same block stream Spectrum's own Hyperliquid nodes consume, from a machine you know the address of.
99.9% uptimeMeasured per calendar month, on gossip reachability to your allowlisted IP from outside our own network. Scheduled maintenance is announced at least 48 hours ahead.
Flat $400/monthPer allowlisted node IP. No usage, bandwidth or egress charges, billed monthly, cancel any time.
Would you rather we ran the node?
Peering is for teams that operate their own Hyperliquid node. If you want Spectrum to run the node and the data services on top of it, that is a dedicated cluster instead.
Everything downstream rides on one peer connection
The public root peer list is the default answer, and it is the reason most self-run nodes drift off the tip. A reserved slot turns the peer you depend on from a lottery into a named machine.
One peer actually feeds your node
A Hyperliquid node lives on the peer connection that feeds it. You inherit that peer's throughput and its uptime. If it is slow, your node is slow. If it drops, your node stops.
Public root peers are a dice roll
The community root peers listed by the gossipRootIps API are shared and oversubscribed, and which one you land on is chance. Addresses change and slots fill, so the peer that worked yesterday may not take you today.
Falling behind is expensive
When your node loses its peer it stops advancing, falls behind the tip and restarts sync. Downstream that reads as stale order books, missing fills and gaps in whatever you write to disk.
A reserved slot removes the guesswork
Your node IP sits in reserved_peer_ips on both Spectrum peers. The slot is held for you, and you know exactly which machine you are pulling from.
Check these before you subscribe
All three are properties of your node and your network, not of ours, so a reserved slot cannot make up for a missing one.
Ports 4001 and 4002 open to the internet
Hyperliquid gossips over both. A node that does not accept inbound connections on these ports is deprioritized by every peer in the network, ours included.
A static public egress IP
Spectrum allowlists that exact address. Dynamic IPs and large NAT ranges cannot be reserved.
hl-visor in non-validator mode
Peering covers non-validator nodes on Mainnet. You keep running and upgrading the visor yourself.
From your IP to a held slot in one business day
Allowlisting is a manual change on our side, so expect your peer addresses within one business day of sending us your node IP. Once you subscribe we open a shared Telegram or Slack Connect channel so you can reach the team directly.
1
Request access
Email [email protected] with the public IP of each node you want to peer and the region it runs in.
2
Receive your peer addresses
We add each node IP to reserved_peer_ips on both peers and send you the two addresses to point at.
3
Point your node at them
Write the peer addresses we send you into ~/override_gossip_config.json, replacing the examples below.
~/override_gossip_config.json
{
"root_node_ips": [{ "Ip": "203.0.113.10" }, { "Ip": "203.0.113.11" }],
"try_new_peers": false,
"chain": "Mainnet",
"reserved_peer_ips": []
}4
Restart and confirm
Restart hl-visor, then check that the node keeps applying blocks. If it stops advancing, or the visor logs show repeated peer disconnects, send us the logs and we will check the reserved slot from the peer side.
Peering
$400per month, per node IP
- 2 Spectrum peers to connect to, single region
- A reserved slot on both, per allowlisted node IP
- Full gossip, the stream our own nodes run on
- 99.9% uptime, measured per calendar month
- No usage, bandwidth or egress charges
- Monthly billing, cancel any time
Request peering access
Every subscriber gets the same peers, the same configuration and the same priority. Our own infrastructure runs on them, on the same terms.
What peering does not include
- A managed Hyperliquid node. You run, upgrade and monitor hl-visor yourself.
- RPC endpoints or data APIs on top of the peer. Peering carries gossip and nothing else.
- A historical backfill. A node that is far behind still has to catch up on its own.
- Validator peering. The offer covers non-validator nodes on Mainnet.
- Tuning of your own hardware, disks or network path.
Start peering
Give your Hyperliquid node a peer it can count on
Send us the public IP of the node you run and the region it runs in. We reserve it on both peers and send you the addresses to point at.
Flat $400 per month, per node IP. Cancel any time.