Quick answer
How do I set up WireGuard IPv6?
WireGuard IPv6 takes 5 changes on a TunnelVPS server, which has one IPv4 and a routed IPv6 /64. Add an IPv6 address to the server's [Interface], one IPv6 /128 per peer, IPv6 forwarding in sysctl, and an ip6tables NAT rule or routed addresses. On each client, add ::/0 to AllowedIPs.
In brief
- Server: an IPv6 tunnel address next to
10.8.0.1/24inAddress. - Peers: one IPv6
/128each, added to theirAllowedIPson the server. - Forwarding:
net.ipv6.conf.all.forwarding=1, with an eye on router advertisements. - Clients: an IPv6
Addressand::/0inAllowedIPs.
What are the 5 steps to add IPv6 to the tunnel?
| Step | Where | What to do |
|---|---|---|
| 1 | Server | Check the server's IPv6 address and route with ip -6 addr and ip -6 route. |
| 2 | Server | Add an IPv6 tunnel address to Address in wg0.conf. |
| 3 | Server | Give each [Peer] an IPv6 /128 in AllowedIPs. |
| 4 | Server | Turn on IPv6 forwarding and add the ip6tables NAT lines. |
| 5 | Clients | Add the IPv6 address and ::/0 to each client file, then reconnect. |
Why add IPv6 to a WireGuard tunnel?
To reach IPv6-only services and to stop IPv6 traffic from leaving outside the tunnel. A phone on a dual-stack network sends IPv6 directly if the tunnel carries only IPv4. That is a leak, even with 0.0.0.0/0 set.
Every server has one dedicated IPv4 address and a routed IPv6 /64. Each plan states its monthly outbound traffic in TB, and going past it never adds an automatic charge. So the server side is ready; WireGuard IPv6 is only a matter of configuration. The DNS leak guide shows how to test for this kind of leak.
Which IPv6 addresses should the tunnel use?
Two options: private ULA addresses with NAT, or addresses taken from the server's own /64. The first mirrors the IPv4 setup and makes all devices share the server's IPv6 address. The second gives each device a public IPv6 address of its own.
With NAT, pick a random prefix from the unique local range fd00::/8; the samples use fd12:3456:789a::/64 as a stand-in. For routed addresses, the samples below use 2001:db8:1234::/64, the documentation prefix, in place of your real /64. TunnelVPS read RFC 4193 on 16 November 2026: the 40-bit Global ID of a local prefix must be pseudo-random, never sequential or well known.
[Interface]
Address = 10.8.0.1/24, fd12:3456:789a::1/64
ListenPort = 51820
PrivateKey = <SERVER_PRIVATE_KEY>
PostUp = iptables -t nat -A POSTROUTING -o eth0 -j MASQUERADE; ip6tables -t nat -A POSTROUTING -o eth0 -j MASQUERADE
PostDown = iptables -t nat -D POSTROUTING -o eth0 -j MASQUERADE; ip6tables -t nat -D POSTROUTING -o eth0 -j MASQUERADE
[Peer]
# phone
PublicKey = <PHONE_PUBLIC_KEY>
AllowedIPs = 10.8.0.2/32, fd12:3456:789a::2/128How do you turn on IPv6 forwarding safely?
Set net.ipv6.conf.all.forwarding=1 in a sysctl file, but check the server's IPv6 route first. The kernel docs say that with forwarding on, router advertisements are accepted only if accept_ra is 2. A server that learns its route that way could lose IPv6.
Run ip -6 route before and after. If the default route vanishes, set accept_ra to 2 on the internet interface. The IP forwarding guide covers the sysctl file in full.
ip -6 addr
ip -6 route
echo "net.ipv6.conf.all.forwarding=1" > /etc/sysctl.d/98-wireguard-ipv6.conf
sysctl --system
ip -6 routeWhat changes in each client file?
An IPv6 address in [Interface] and ::/0 in AllowedIPs. The address must match the peer's IPv6 on the server. Reconnect the tunnel after the edit.
The add a device guide builds a complete client file. After this change, the device should show the server's IPv6 address on any page that displays your public IP. That confirms WireGuard IPv6 works end to end.
[Interface]
PrivateKey = <PHONE_PRIVATE_KEY>
Address = 10.8.0.2/32, fd12:3456:789a::2/128
DNS = <DNS_RESOLVER_IP>
[Peer]
PublicKey = <SERVER_PUBLIC_KEY>
Endpoint = 203.0.113.10:51820
AllowedIPs = 0.0.0.0/0, ::/0When are routed addresses from the /64 better?
When a device must be reachable on its own public IPv6, for example to host a service. Give the peer an address such as 2001:db8:1234::2/128 and skip the ip6tables NAT line. The prefix must be routed to the server, which is how TunnelVPS describes its /64.
For privacy, NAT is usually the better default: every device leaves through one address. A personal VPN hides traffic from the local network and the internet provider, not from the sites visited: every connection leaves through one IP address rented by one account, so it is a private tunnel, not anonymity.
WireGuard IPv6 questions
Does every TunnelVPS plan include IPv6?
Yes. Every plan has one dedicated IPv4 address and a routed IPv6 /64, at the same price in all six cities.
Can WireGuard IPv6 work with an IPv4-only endpoint?
Yes. The tunnel can carry IPv6 inside while devices connect to the server's IPv4 address.
What is the best WireGuard IPv6 setup in 2026 for a household?
ULA addresses with NAT: simple, private, and matching the IPv4 setup.
Why do IPv6 sites still fail after the change?
Usually IPv6 forwarding is off, the ip6tables rule is missing, or the client lacks ::/0 in AllowedIPs.
Does UFW need changes for IPv6?
Route and port rules must cover IPv6 too. Check ufw status verbose after adding IPv6 to the tunnel.
Verdict
- Add IPv6 to the tunnel rather than leave it outside: it closes a real leak.
- ULA with NAT is the simple default; routed /64 addresses suit hosted services.
Sources (4)
- IP Sysctl (forwarding, accept_ra) — kernel.org (read 2026-11-16)
- wg-quick(8) manual page — man7.org (read 2026-11-16)
- Unique Local IPv6 Unicast Addresses (RFC 4193) — IETF (read 2026-11-16)
- IPv6 Address Prefix Reserved for Documentation (RFC 3849) — IETF (July 2004)
IPv4 and a /64 on every plan
Home: 3 shared vCPU, 4 GB RAM, 60 GB NVMe, 5 TB a month, one IPv4 and an IPv6 /64, at $6.37 a month on 12 months.
- Built by a person, 1 hour at most after confirmation
- Lost days credited back if a server cannot be delivered
- Pay in BTC, XMR, USDT, USDC, ETH, SOL or LTC
- DomaineGoat LLC · Panama