About IPv6 ULA Generator
IPv6 Unique Local Addresses (ULA, fc00::/7) are the IPv6 equivalent of private addresses — for internal use and never routed on the public Internet. Per RFC 4193, a proper ULA prefix sits under fd00::/8 and uses a random 40-bit Global ID so different sites almost never collide. This tool generates spec-compliant /48 ULA prefixes and splits /64 subnets locally in your browser with no network access — ideal for planning internal IPv6 addressing. When planning internal addressing, the public block allocated by a carrier or registry rarely suits an internal topology, while reusing private ranges risks colliding with a partner during interconnection. A locally unique address block is designed for exactly this: it is identified by a fixed prefix and is valid only inside the organisation. So it never collides with public space, which makes it a good fit for intranets without public requirements, container networks and lab environments. Generation takes the fixed prefix and appends a randomly generated forty-bit identifier to form the complete prefix. Because the identifier space is vast, the chance of two organisations generating the same block is negligible. The important nuance is the double edge of that randomness: it guarantees no collision, and it also means regenerating without recording the result yields a completely different block, so the planning process must include a registration step. Practical notes: Allocate a distinct block per site or purpose rather than sharing one large block company-wide, which keeps boundaries clear for route aggregation and access control later. Decide the subnet length at the same time, commonly one subnet per physical location or availability zone, to avoid renumbering later. Since the addresses are only valid internally, keep them out of externally published configuration and documentation, which would both expose internal structure and create external dependencies on an address that cannot be routed anyway. Limits: Local addresses never appear in public routing, so if external access to an internal service is ever needed, it must go through a tunnel or gateway translation rather than an expectation that the carrier will route it; plan the remote access approach up front. The random block also carries no semantics, so do not try to infer location or environment from an address, and keep one address plan as the single source of truth. Finally, although collisions are extremely unlikely, verify both sides' blocks when merging networks or interconnecting with a partner, and translate addresses if needed, or the routing table ends up with overlapping prefixes that are painful to debug. Misuse centres on two points: not registering the random block, and treating it as routable. A generated block cannot collide with others, yet without a record the same block cannot be reproduced and regenerating yields a completely new prefix, so registration is a required step rather than optional. The other misuse is putting local addresses into externally published configuration, which exposes internal structure and creates external dependencies on an unroutable address.
How It Works Internally
The prefix is a fixed part plus a randomly generated forty-bit identifier. That randomness cuts both ways: it guarantees no collision with other organisations and it means an unrecorded value cannot be reproduced, so registration is mandatory or regenerating yields something entirely different. Such addresses never appear in public routing, so external access needs a tunnel or gateway translation.
Generate a ULA prefix onceA fixed 48-bit prefix (fixed part plus 40 random bits); regenerating gives a different value, so registration is requiredMeasured on This Page
| Structure | All five follow fd + 40-bit global ID + 16-bit subnet at /64, e.g. fde4:7b25:92b2:8352::/64 |
|---|---|
| RFC 4193 placement | All inside fd00::/8 (locally assigned ULA) with prefix length 64 |
| Randomness | Seven bytes from crypto.getRandomValues; all five distinct |
| Cap | At most 50 per batch (a larger count is clamped) |
| Copy | Single and copy-all both work |
ULA is the private range that never routes publicly and cannot collide with provider allocations, which makes it good for stable home or intranet addressing — unlike NAT-ed 192.168.x.x it survives a broadband reconnect, and unlike fe80:: link-local it carries no interface identifier. Note the L bit (bit 8) is 1 in fd00::/8 meaning locally assigned, while fc00::/8 is reserved for future registries, so self-generated prefixes must start with fd. A generated address is only probably unique, so generate once and record it.
How to Use
- Open "IPv6 ULA Generator"
- Set how many prefixes you need, up to 50
- Generate a list of ULA prefixes under fd00::/8
- Each uses a random 40-bit Global ID per RFC 4193
- Copy one row, or copy all prefixes at once
Use Cases
- Internal addressing — Generate a stable /48 ULA prefix for a home or company network, independent of the ISP.
- Subnet planning — Carve multiple /64 subnets under a /48 for different VLANs or segments.
- Site interconnect — Produce low-collision ULAs for multi-site VPNs to avoid clashes when merging.
- Lab environments — Use local addresses that never leak to the public Internet when building an IPv6 lab.
- Dual-stack transition — Carry internal services over ULA while public traffic uses global addresses.
- Container networks — Generate a /64 per Docker or Kubernetes pod range so each tenant gets a unique local segment.
- Collision check — Compare generated ULAs against your existing site prefixes to confirm no overlap before deployment.
FAQ
How does ULA differ from link-local (fe80::)?
fe80::/10 link-local addresses are valid only on a single link and cannot cross routers; ULA (fd00::/8) routes across an entire site between subnets but, like link-local, is never routed on the public Internet.
Why do generated prefixes start with fd?
ULA spans fc00::/7; with the L bit set to 1 (i.e., fd00::/8), the Global ID is locally and randomly assigned, as RFC 4193 recommends. fc00::/8 is reserved for future central allocation and is not yet in use.
Why must the Global ID be random?
A random 40-bit Global ID makes collisions between different organizations' ULA prefixes extremely unlikely, so addresses rarely clash when sites merge or VPNs interconnect.
How many /64s fit in one /48?
From /48 to /64 there are 16 subnet bits, allowing 65,536 /64 subnets — ample for most internal networks.
Can a ULA reach the public Internet directly?
No. ULAs are designed not to be routed publicly; Internet access still needs a global unicast address (GUA) or NAT/NPT translation.
Can a generated ULA prefix be routed publicly?
No, and that is the design intent. Locally unique addresses are valid only inside the organisation and never appear in public routing, so any external access to an internal service needs a tunnel or gateway translation rather than an expectation that the carrier will route it. Settle the remote access approach during planning, or you discover the missing entry point after go-live.
Why is a regenerated prefix completely different from the previous one?
Because the random portion is drawn afresh each time and the tool keeps no history. That randomness is what prevents collisions with other organisations, but without a record the same block cannot be reproduced. The planning process must include registration: record the value in the address plan under version control immediately, and take it from there afterwards instead of regenerating.
Can ULA addresses be used on the public internet?
No. ULA (fd00::/8 in practice) routes only within a site and must not be announced publicly. It suits internal static addressing without DHCP; internet access still needs global unicast addresses via NAT or direct assignment.