Fibmesh Public IPs / How it works

One address.
Four decisions.

Separate the allocation, its delivery, the traffic it carries and the rules for incoming access. Each choice answers a different question.

Follow the decisions Public addressing guide · Confirm platform and deployment support during assisted setup.

Available with assisted setup. Confirm the supported platform, serving location, inventory and traffic policy for your deployment.

01 / Keep the choices independent

Start with the address.
Decide how to use it.

A public allocation is separate from a device’s identity and the private addresses it uses inside Networks. A private Networks network is optional; account identity and entitlement are still required.

  1. 01 / Allocation

    Which address?

    A single target or a separately allocated block. Choose the address families your clients and destinations need.

  2. 02 / Delivery

    Where does it arrive?

    Route the address directly to a supported endpoint, or forward permitted traffic through a gateway to LAN equipment.

  3. 03 / Traffic

    Which connections use it?

    Choose incoming access, selected outgoing traffic or full tunnel. Give replies their own correct return path.

  4. 04 / Firewall

    Who may come in?

    Keep new incoming connections closed until you explicitly permit the required callers, protocols and ports.

Single target

A device or gateway address.

Dual-stack gives a supported target IPv4 and IPv6. IPv6-only needs compatible clients and destinations; it does not automatically include an IPv4 address or translation.

Infrastructure

A separately allocated block.

Allocate IPv4 and IPv6 blocks separately. IPv6 prefixes range from /48 through /64. Route the block to your infrastructure, then assign addresses or downstream subnets to workloads.

Explore Routed Subnets →

02 / Delivery paths

The same public destination.
Two different arrivals.

Your ISP transports the outside of the encrypted tunnel. The assigned public address identifies the traffic inside it. ISP NAT and public-address translation are separate questions.

Direct routed delivery

The address reaches the host.

  1. Public destination203.0.113.24
  2. WireGuard tunnelFibmesh → host
  3. Endpoint destination203.0.113.24

Packets retain the assigned public destination all the way to the supported endpoint. No destination NAT is required there. A fully routed address can still have a firewall.

Direct delivery to a device →
Gateway forwarding

A rule selects the LAN target.

  1. Public destination203.0.113.24
  2. GatewayPermitted forwarding
  3. Private LAN target10.20.0.10

The gateway translates permitted traffic to an explicit private target. LAN-side source NAT can preserve the equipment’s existing configuration, but the target may see the gateway’s address instead of the original client’s.

Forwarding through a gateway →
The tunnel is transport.

WireGuard is currently the only supported tunnel. Confirm a Fibmesh-supported client or gateway; generic WireGuard capability does not establish support. The ISP may use NAT or CGNAT, provided the outbound WireGuard UDP tunnel can be established and maintained. Compatibility is confirmed for the deployment. Manual WireGuard profiles are Compatibility Mode, rather than full managed functionality.

Documentation addresses illustrate the paths. Inbound-only delivery can leave ordinary new outgoing traffic on the ISP, while service replies return through Fibmesh. NAT alone does not choose that return route.

03 / Traffic policy

Choose the traffic.
Check the source.

Incoming access and outgoing routing are independent. Start with a preset, then verify the actual paths at the supported endpoint or routed gateway.

Explore a traffic preset Illustration · no settings are changed

Traffic preset
Internet clientFibmesh + inbound rulesYour service

Let the intended connections in.

Permitted incoming connections use Fibmesh, with a matching return path for their replies. New, unrelated outgoing connections keep their existing ISP route.

Incoming: allowed by rulesOrdinary outgoing: existing ISP
Your deviceAssigned Fibmesh sourceSelected destinations

Use your public address where it matters.

Chosen destinations see the assigned Fibmesh source address. Other destinations use the existing ISP. Replies to your connections are allowed.

Add incoming access too

Enable separate inbound rules to host a service while making selected outgoing calls from the same address. Incoming access is blocked by default in the selected-outbound preset.

Incoming: blocked by defaultOther outgoing: existing ISP
Your deviceFibmesh tunnelEligible internet traffic

Make Fibmesh the internet path.

Eligible internet connections use Fibmesh. Incoming access remains a separate firewall choice. Confirm both address families, disconnect behaviour and local or platform exceptions.

Incoming: independent rulesIPv4 + IPv6: explicit policy
All three presets at a glance
PresetNew incoming connectionsNew outgoing connectionsTraffic left on the ISP
Inbound accessPermitted by incoming rules; replies return through Fibmesh.Ordinary new outgoing traffic keeps its existing route.Unrelated outgoing connections.
Selected outboundBlocked by default; enable incoming access separately. Replies to initiated connections remain allowed.Agreed destinations use Fibmesh and the assigned source address.Destinations outside the selection.
Full tunnelStill controlled by independent incoming rules.Eligible internet traffic uses Fibmesh, with explicit IPv4 and IPv6 policy.Documented local and platform exceptions; any deliberately bypassed family must be labelled.

A path and a source.

Selected destinations are agreed IP addresses and CIDR ranges. Select the assigned public source as well as the route: choosing the tunnel path alone is insufficient. Domain or application selectors require separate client support; this guide does not establish those controls as implemented.

Incoming and outgoing can coexist.

Host a service and call a partner API from the same address with inbound plus selected outbound. Blocking new incoming connections must still allow replies to connections you started.

Does a plug-in connector route every device at a site?

Joining the LAN does not redirect its outgoing traffic. Quick Connect forwards selected incoming traffic to a configured target, or through per-port mappings on supported software. New outgoing connections must actually traverse the gateway through routing changes or a protected downstream network. A one-port connector does not isolate the LAN or provide independent recovery when the ISP fails. Edge hardware remains a development/pilot deployment option; confirm the supported software and address-family behaviour for your gateway.

The explorer is an illustration. Its native controls change no settings. Confirm available client or gateway controls during assisted setup.

04 / Incoming boundary

An address is a destination.
Permission is a rule.

Full tunnel does not open incoming access. An outgoing allowlist does not publish a service.

Understand the firewall boundary →
Managed inbound / starts closed

Open only the intended service.

  • Permit the required caller source ranges, protocols and ports.
  • Confirm protocol support and reserved platform management ports.
  • Keep the host firewall and application authentication. Handle essential network control traffic correctly.

Fully open exposure, where supported, requires explicit confirmation as an advanced choice. A public IP and an IP allowlist are not user identity or application authentication.

05 / Families & failure

Agree the exceptions.
Test the interruption.

“Full tunnel” describes the intended traffic policy. Confirm what the supported platform can enforce for both address families and during tunnel loss.

IPv6-only + IPv4
Choose an explicit IPv4 policy: block it, use a separately supported translation service, or deliberately leave it on the ISP. ISP bypass is split-family routing and must be labelled. Apply the same explicit decision to any unsupported family.
Disconnect behaviour
Agree and verify whether traffic assigned to Fibmesh is blocked by a kill switch when the tunnel fails. If fallback is allowed, destinations may see the ISP’s source address. Confirm recovery before relying on an external allowlist; no blanket disconnect protection is implied.
Platform exceptions
Mobile OS-managed VPNs have routing and lifecycle constraints. Confirm local-network and system-traffic exceptions. Full tunnel does not promise that every system packet follows the same path on every platform.
Encryption boundary
The tunnel protects the endpoint-to-Fibmesh leg and ends at the Fibmesh peer. Use TLS or other application encryption beyond that point, together with application authentication.

06 / Domains & applications

Name the address.
Secure the application.

Point a domain you control to the assigned IPv4 address with an A record, or to its IPv6 address with an AAAA record. The application or a reverse proxy you operate supplies HTTPS and its certificate.

Documentation example / IPv4
files.studio.exampleA record →203.0.113.24

DNS names the destination. Your service owns its listener, authentication and TLS.

A Public IP does not automatically create a Fibmesh subdomain, provide a certificate or choose applications by hostname. A DNS record also does not select a non-standard service port.

Choose Public IPs for an address with supported TCP or UDP services. For one HTTPS link to a web app, consider Publish.

Compare a public link with a Public IP →

07 / Before handover

Accept the real paths.
Include the failure case.

A tunnel handshake is not an application test. Check what can connect, what cannot, which source appears and what happens when delivery stops.

  1. 01 / Acceptance

    Incoming permission

    Reach the actual application from a permitted external caller. Confirm a denied caller cannot reach it.

  2. 02 / Acceptance

    Outgoing identity

    Test selected and unselected destinations. Check the source address each destination actually sees.

  3. 03 / Acceptance

    Address families

    Test IPv4 and IPv6 separately. Confirm any unsupported family is blocked, translated by a supported service or clearly labelled as an ISP bypass.

  4. 04 / Acceptance

    Return & recovery

    Check gateway return routing and MTU with a larger transfer. Interrupt the tunnel, verify the agreed block or fallback, then test recovery.

Reservation and delivery have different lifecycles.

Pause
Withdraw delivery while retaining the reserved address. Verify the effect on traffic and recovery.
Release
Give up the address. Update DNS, external allowlists and other references before release.
Region
A preference change does not move the same address. Retention does not promise seamless sessions, roaming, automatic failover or high availability.

Next / Your traffic requirements

Bring the connections.
Agree the boundaries.

Tell us the target, address families, callers and outgoing destinations. We’ll confirm the supported deployment, serving location and traffic policy during assisted setup.