Peering guide

Five minutes from login to feeding. We run receive-only eBGP — we accept every route you send, announce nothing back, and deduplicate so identical views don't cost anyone storage.

The deal

eBGP, multihop welcome — TTL doesn't matter, we're not in your path.
We never send UPDATEs. Your router sees a session that receives everything and advertises zero routes — safest possible neighbor.
v4 session, v6 session, dual-stack, or cross-AFI (v6 NLRI over a v4 session) — your choice when you create the session.
ADD-PATH receive negotiated on both AFIs — run addpath send and we'll take every path, not just best.
Route-refresh + enhanced route-refresh supported; EOR marks your table transfer complete.
We accept full tables, partial views, RIB-in dumps of just your downstream — whatever you've got. More vantage = better map.
Max-prefix ceilings exist for our sanity but they're silent — your session stays healthy either way.
MD5 is supported but optional; sessions are allowlisted by source IP anyway.

1 · Get affiliated

Sign in with PeeringDB. The networks scope tells us which ASNs are yours — if your AS isn't listed on your PeeringDB profile, affiliate it there first. That verification is what makes our peer data trustworthy downstream.

2 · Create a session

From /sessions: pick a collector, a transport (v4 or v6 — must match your router IP), and what you'll send. The panel shows the service address to peer with. Run two collectors if you can — redundancy is nearly free.

This swarm's service addresses

IPv4 → 23.135.148.5 · IPv6 → 2602:f73b:400::5 · remote AS 65000

3 · Configure your side

Receive-only means your export policy is "send everything to us, accept nothing back" — or just send and let our silence do the filtering.

pathvector

peers:
  bgpath:
    asn: 65000
    neighbors:
      - 23.135.148.5
      - 2602:f73b:400::5
    multihop: true
    add-path-tx: true        # send us every path, not just best
    export-default: false
    # pathvector's import policy already drops non-unicast junk;
    # we announce nothing anyway.

BIRD 2

protocol bgp bgpath4 {
  local as YOUR_ASN;
  neighbor 23.135.148.5 as 65000 multihop;
  ipv4 { export all; import none; add paths tx; };
}
protocol bgp bgpath6 {
  local as YOUR_ASN;
  neighbor 2602:f73b:400::5 as 65000 multihop;
  ipv6 { export all; import none; add paths tx; };
}

FRR

router bgp YOUR_ASN
 neighbor 23.135.148.5 remote-as 65000
 neighbor 23.135.148.5 ebgp-multihop 16
 address-family ipv4 unicast
  neighbor 23.135.148.5 activate
  neighbor 23.135.148.5 addpath-tx-all-paths
 exit-address-family
 neighbor 2602:f73b:400::5 remote-as 65000
 neighbor 2602:f73b:400::5 ebgp-multihop 16
 address-family ipv6 unicast
  neighbor 2602:f73b:400::5 activate
  neighbor 2602:f73b:400::5 addpath-tx-all-paths
 exit-address-family

Juniper

set protocols bgp group bgpath type external
set protocols bgp group bgpath multihop ttl 16
set protocols bgp group bgpath peer-as 65000
set protocols bgp group bgpath neighbor 23.135.148.5
set protocols bgp group bgpath neighbor 2602:f73b:400::5
set protocols bgp group bgpath family inet unicast add-path send path-count 4
set protocols bgp group bgpath family inet6 unicast add-path send path-count 4
set protocols bgp group bgpath export send-everything

Arista EOS

router bgp YOUR_ASN
   neighbor 23.135.148.5 remote-as 65000
   neighbor 23.135.148.5 ebgp-multihop 16
   neighbor 23.135.148.5 send-community
   address-family ipv4
     neighbor 23.135.148.5 activate
     neighbor 23.135.148.5 additional-paths send any
   address-family ipv6
     neighbor 2602:f73b:400::5 activate
     neighbor 2602:f73b:400::5 additional-paths send any

What happens next

Your dashboard flips pending → established the moment the session comes up — live FSM events, no refresh needed.
Prefix counts and update rates stream in as they happen.
Identical re-announcements are absorbed at the collector edge — a resync doesn't flood the pipeline, and your feed's unique contribution is what's stored.
Everything lands in public MRT archives and the live stream — the data you feed is the data everyone gets.

IXP operators

Want a collector on your fabric? An IXP node is the same agent on your peering LAN — members peer directly (TTL=1 works), your IX gets credited as the collector's host. See /sponsors.

Consumers

Don't feed? Take instead: MRT archives, /api/v1 for queries, and the public real-time stream (RIS Live-compatible JSON) for tooling like bgpkit and ARTEMIS.