plan
step 1> PrefixService.AllocateNextFreeSubnet
pool10.42.0.0/1665,536 addresses
size/2662 usable hosts
policyfirst free
dispatchintent, realityin parallel
waiting on both
The agent-native control plane for DDI and DCIM. One live twin of your estate, on open standards, with a deterministic Rust math core underneath. Signed lease ingestion and tenant isolation work today; the agents that would act on the twin are in development.
$ omnitwin agent launch scout --subnet 10.0.0.0/16
✓ sweeping 65,534 addresses [] 100%
✓ 1,284 hosts found · 47 undocumented
$ omnitwin drift detect --scope production
✓ comparing intent to reality [] 100%
! 12 records diverged from intent
$ omnitwin reconcile --plan drift-8f2a --apply
✓ applying 12 changes [] 100%
✓ twin aligned · 0 conflicts · audit logged
$ _ Speaks the protocols you already run
Discovery, allocation, drift, pre-flight and audit all read and write the same live model of your estate, so an answer from one is an answer the others already agree with. Each surface below says plainly whether it ships today.
A sweep reads the estate directly rather than trusting the record of it, then reports the three numbers that matter: what answered, what the IPAM already knew about, and the gap between them.
In development$ omnitwin discover --subnet 10.42.0.0/16 --probe icmp,snmp,arp
✓ sweeping 65,534 addresses [] 100%
responding 1,284
already in IPAM 1,237
undocumented 47
$ omnitwin discover adopt --undocumented --dry-run
✓ 47 hosts would be written to the twin
✓ 0 collisions with existing records
$ _ The math core walks the pool in order and returns the first block that fits. No spreadsheet, no reviewer, no second answer on a second run. This pane prints what the arithmetic returns for the pool below, computed at build.
Math core ships today$ omnitwin alloc next --pool 10.42.0.0/16 --size /26
policy first-free · deterministic · no agent in the path
held 704 addresses across 4 reservations
examined 11 candidate blocks in order
✓ 10.42.2.128/26 free · 62 usable hosts
$ omnitwin alloc reserve 10.42.2.128/26 --for k8s-pods
✓ reserved · written to intent · audit logged
$ _ Two stores answer the same question: the one that holds what you meant, and the one that holds what the network is doing. Drift is every record where those two disagree, listed rather than summarised.
In development$ omnitwin drift detect --scope production
✓ comparing intent to reality [] 100%
! 12 of 41,908 records diverged
A app-7.prod-east 10.42.9.14 reality 10.42.9.40
A cache-2.prod-east 10.42.9.21 reality absent
PTR 10.42.7.0/24 present reality absent
… 9 more
$ omnitwin drift reconcile --plan drift-8f2a
! 12 changes staged · awaiting one approval
$ _ A proposed change runs against a branch of the twin in memory before production sees it. Overlapping CIDRs, exhausted scopes and broken delegation are caught by hard rules, so the push fails instead of the estate.
Ships today$ omnitwin apply plan.tf --simulate
cloning twin branch sim/8f2a · in memory · nothing shipped
✓ vlan 340 delegation resolves
✓ dhcp scope headroom 38%
✓ zone prod-east.internal signs clean
✗ 10.128.0.0/20 overlaps sn_01HX7
blocked before production · 1 conflict
discarding branch sim/8f2a
exit 1
$ _ Every lease arrives with a signature from the server that issued it and is checked before it reaches the twin. The tenant it belongs to is enforced on the same path, not after the fact.
Signed ingestion ships today$ omnitwin audit tail --tenant acme --limit 4
14:02:11 lease.ingest dhcp-02 sig ✓ ed25519
14:02:11 lease.commit 10.42.7.31 ttl 3600
14:03:04 subnet.reserve 10.42.2.128/26 agent:maya
14:03:04 policy.check tenant-isolation pass
✓ 4 records · 4 signatures verified · 0 unsigned accepted
✓ every write attributable to a key, not a username
$ _ A request never lands in one place. Plan reads it and dispatches, the intent store and the reality store answer at the same time, and the math core waits for both before it returns an address.
> PrefixService.AllocateNextFreeSubnet
pool10.42.0.0/1665,536 addresses
size/2662 usable hosts
policyfirst free
dispatchintent, realityin parallel
waiting on both
> reservations held in 10.42.0.0/16
10.42.0.0/24256 addresses
10.42.1.0/24256 addresses
10.42.2.0/25128 addresses
10.42.3.0/2664 addresses
4 blocks held704 addresses
> observed state for 10.42.0.0/16
blocks cross-checked4
in-flight leasesnone observed
conflicting edgesnone observed
drift against intentnone observed
clear to allocate
> first_free(pool, /26, held)
0110.42.0.0/26heldby 10.42.0.0/24
..8 moreheldby 3 reservations
1010.42.2.64/26heldby 10.42.2.0/25
1110.42.2.128/26free
→ 10.42.2.128/26
A design preview of the trace, not a recorded session. The gRPC method and the arithmetic are real and every address is computed; the agents that would drive this path are in development.
The design puts agents beside the data rather than calling out to it: persistent state, identity-bound access, no context re-hydration per call. Agent interfaces are in development and no agent is wired yet.
Designed for the engineer with a credit card. A $25 entry tier that is actually useful, and a self-serve path all the way up.
Drift is the gap between what you intended and what the network is doing. Surfacing it, and reconciling it on one approval, are in development. The what-if check that tests a change before production ships today.
$1,250/month against legacy platforms billing six figures a year. Same capability, different foundations.
Passive telemetry only sees a device when it decides to talk. Scout is designed to sweep the subnet directly over ICMP, ARP and SNMP, so silent bare-metal, static assignments and unallocated hardware all surface. Device discovery is not built yet, and the session below is a design preview.
$ omnitwin agent launch scout --subnet 10.20.0.0/16
probing icmp · arp · snmp
✓ 10.20.4.17 dell poweredge r760 undocumented
✓ 10.20.4.18 cisco nexus 9300 matches twin
✓ 10.20.9.204 silent bare-metal undocumented
1,284 hosts · 47 undocumented · 12 stale $ omnitwin apply plan.tf --simulate
cloning twin branch sim/8f2a
✓ vlan 340 delegation resolves
✓ dhcp scope headroom 38%
✗ 10.128.0.0/20 overlaps sn_01HX7
blocked before production. 1 conflict. The what-if check tests a proposed change against your twin, in memory, before production sees it. Overlapping CIDRs, exhausted scopes and broken zone delegation get caught by hard mathematical rules, not by a reviewer. This one ships today, as ordinary code rather than as an agent. The command below is a design preview: the tenant CLI is still in development.
When operational state diverges from the source of truth, the canvas is designed to show you, and to reconcile on one approval with a record of who changed what. Drift detection is in development.
Built different from the foundations up, not bolted onto a record store.
OmniTwin sits alongside the DNS, DHCP and CMDB systems you already run. No rip and replace: your infrastructure stays yours and gains a live twin of itself.
Agents catch drift the instant it happens and stage the reconcile. No stale spreadsheets, no 3am page for a mismatch that was preventable.
Changes run through a sandboxed twin before they reach production. Mathematical proof that a change cannot cause an overlap, a leak or a policy violation.
Legacy DDI platforms bill six figures a year on rigid licences. OmniTwin starts at $25 a month with the same enterprise capability.
Open-standards DDI is not just architecture. It is a global repository of networking practice, stress-tested by thousands of engineers who face the same complexity you do.
Unlike proprietary tools built by generalist vendors, OmniTwin was forged in the trenches of real network operations, with agents that act on their own.
DDI should not be a spreadsheet your team dreads updating. It should be a living twin that discovers, detects drift and reconciles.
Every layer is documented, from the architecture down to the endpoints.
Dual-database design with Rust math validation, a Go control plane and graph-powered topology awareness.
Explore architectureRESTful endpoints for every resource, with a full OpenAPI spec, auth, pagination and webhooks.
Read the docsRun queries against a live sandbox. Test CIDR math, graph traversals and agent workflows interactively.
Open playgroundWe protect it with the rigour you would expect from the systems it manages.
End-to-end encryption, zero-trust networking, WAF protection and continuous vulnerability monitoring across all infrastructure.
View security practicesBuilt to support SOC 2 Type II and ISO 27001 controls, with GDPR and HIPAA-eligible data handling. Immutable audit logs and data residency controls.
View our compliance positionData ownership guarantees, responsible agent policies, transparent operations and business continuity commitments.
Visit trust centreThe beta is now open. Every signup gets an access code: it holds your seat in a cohort and puts you first in line for launch pricing and unlocked features.
Takes you to the beta program page to claim your code. By signing up you agree to our terms and privacy policy.