Frequently asked
questions.
The questions we hear most. If yours isn't here, the answer is one email away.
General
- Teams migrating from legacy DDI vendors and want a clean, open-standards foundation.
- Organizations where undocumented network changes create compliance or operational risk.
- Platform teams building infrastructure automation and need a DDI API they can actually compose with.
- Any environment where the gap between the documented network and the real network is a known problem.
What is OmniTwin?
OmniTwin is an agent-native, graph-first control plane for DDI (DNS, DHCP, and IP Address Management) and DCIM (Data Center Infrastructure Management), built by Kortesa Labs. It is in private beta, which is now open. The design is a live digital twin of your network that agents will eventually observe and act on. The agent interfaces are in development and no agent runs today.
What works today is the foundation: signed DHCP lease ingestion, subnet allocation, bulk DCIM import through an operator CLI, and tenant isolation enforced in the database. The design goal is that the system of record and the system of action are the same thing, rather than agents sitting on top of a database and querying it.
What is DDI?
DDI stands for DNS, DHCP, and IP Address Management - the three foundational control-plane services that every IP network depends on. DNS resolves hostnames to addresses. DHCP assigns addresses dynamically to hosts. IPAM tracks the allocation, assignment, and availability of every address in your space.
Most organizations manage these three services separately, with different tooling and different data models, which creates reconciliation overhead and drift. Integrated DDI platforms manage all three from a unified data model.
Who is OmniTwin for?
OmniTwin is built for network engineers, NetOps teams, and platform engineers who manage IP space at scale and need automation that goes beyond dashboards and manual workflows. Specifically:
What's the difference between OmniTwin and legacy IPAM tools?
Three fundamental differences:
Architecture. Legacy tools are passive records: engineers query them and enter data manually. OmniTwin is built as an active runtime. Today that shows up in the DHCP lease path, which is event-driven and durable: signed lease events are ingested and processed in two stages, and the twin updates as they arrive. Agents that observe, plan, and act on the twin are the design direction and are in development.
Standards. Legacy proprietary platforms use closed data models that create migration friction and API inconsistency. OmniTwin is built on open-standards DDI: RFC-compliant schemas, portable data, no vendor lock-in.
Where automation sits. Most tools bolt an "AI" feature onto an existing IPAM UI. OmniTwin is designed around agents as first-class components of the runtime. Those agents are not wired yet, so treat them as roadmap, not as a feature you can use today.
Pricing
How does pricing work?
OmniTwin pricing is tiered by capacity and deployment model. There are four tiers: Essential ($25/month, 100 IPs), Growth ($1,250/month, 2,500 IPs), Pro (5,000 IPs on shared high-performance nodes) and Scale (dedicated single-tenant, 1,000,000+ IPs). Pro and Scale are quoted per deployment, so talk to sales for a price. Agents are not bundled into the tiers. Agent interfaces are in development, and execution packages for them will be sold separately once available.
Every tier is billed annually. See the pricing page for the full breakdown.
What is the cheapest way to start?
The Essential plan, at $25 a month. It supports up to 100 IP addresses and 100 devices on shared high-density infrastructure, with the DDI and DCIM core, 10,000 API requests a month, 7-day AI telemetry, social and magic link sign-in, and community support.
No tier bundles agents. Agent interfaces are in development, and execution packages for them will be sold separately on top of any plan, including Essential, once they are available.
What happens if I exceed my IP limit?
We don't cut you off mid-operation. When you approach your IP limit you'll receive a warning via the UI and email. You'll have a grace period to either reduce your tracked addresses or upgrade your plan. We'll always alert you before anything is restricted.
What counts as an IP: any IPv4 or IPv6 address tracked in your IPAM twin, whether assigned, reserved, or learned from DHCP leases. Subnets and ranges are not counted individually.
Can I change my plan or arrange custom billing?
Yes. For questions about plans and billing, or if you need a customized plan or arrangement, reach out to our support team at support@kortesalabs.com. They handle sales and billing and will work through the options with you.
Technical
What open standards does OmniTwin support?
OmniTwin is built on RFC-compliant DNS (RFC 1034/1035), DHCP for IPv4 (RFC 2131), DHCPv6 (RFC 3315), and IPv6 addressing architecture (RFC 4291). All data schemas conform to published standards - there are no proprietary object types that create lock-in or migration friction.
This means OmniTwin can interoperate with any RFC-compliant DNS server, DHCP server, or IPAM consumer without custom adapters. Bulk DCIM import ships today through an operator CLI. Agent-driven translation from other vendor formats is in development.
How does the Universal Importer (The Polyglot) work?
The Polyglot is the import and translation agent. It is partly real today: bulk DCIM import ships through an operator CLI, which loads a DCIM bundle into your twin. Any agent-driven import, where The Polyglot reads vendor exports and maps them to OmniTwin's open-standards schema, is in development.
The design is that import runs as a preview first, so you see what will change before committing, and that ambiguous field mappings are surfaced for human review rather than guessed. Treat that review flow as the design, not as a shipped feature.
Is there an API? What does it cover?
Two surfaces work today. An HMAC-signed HTTP webhook ingests DHCP lease events, and a gRPC method allocates the next free subnet inside a tenant's address space. Both are documented in the API reference, which states the status of every endpoint.
A REST API covering prefixes, addresses, DNS zones and records, DHCP scopes and leases, and devices is in development and not callable yet. Authentication is by JWT or API key, exactly one credential per call, and the tenant is taken from the credential rather than from the request body.
What SSO options are supported?
All plans include standard SSO via Google Workspace, Microsoft Entra ID (Azure AD), and GitHub. Pro, Scale and On-Prem additionally support enterprise SAML 2.0 and group-to-role mapping from your identity provider - so access control follows your existing directory without manual assignment. Scale and On-Prem go further with custom RBAC.
SCIM provisioning for automated user lifecycle management is on the Phase 2 roadmap for enterprise plans.
What are OmniTwin agents?
Agent interfaces are in development, and no agent is wired today. The design is for agents to be software components that run against your OmniTwin twin and take defined actions: discovery, translation, validation, planning, or execution. Each would have a specific job, a trigger model, and a defined scope of authority, and not be an AI chat assistant bolted onto a database.
The one exception that ships is the what-if check. It is ordinary code on a tenant route, not an agent, and it evaluates a proposed change without changing anything.
Agents
Can I choose which agents to run?
That is the design, and it is not available yet because no agent is wired today. The plan is for agents to be independently configurable, so you control when they run, which prefixes, VLANs, or zones they cover, and what actions they are permitted to take.
For example, you could enable drift detection on production without running it across dev or lab space. Which agents you can use would depend on your plan.
How does drift detection work?
Drift is the difference between what your OmniTwin twin records and what is observed on the live network: a host in the twin with no live address, an active host with no twin record, a DNS record pointing to a decommissioned address.
Detection is in development and is not wired today. It is designed as Truth Seeker comparing the twin against what Scout observes, and Scout, the discovery agent, is not built yet. Until it is, OmniTwin does not detect drift for you.
The guarantee that matters is already part of the design: resolving drift requires human approval. It is not automatic.
What is pre-flight simulation and why does it matter?
Pre-flight simulation checks a proposed change, such as a new prefix, DNS record, or DHCP scope, against your twin before anything touches the live network. The what-if check ships today as ordinary code on a tenant route. It is not an agent, and it changes nothing.
Making pre-flight a mandatory gate for every change, run by The Safety Net agent, is the design direction and is in development.
This matters because DNS and DHCP changes can cascade quickly: a misconfigured scope can affect hundreds of hosts, and a bad SOA record can break zone delegation. Checking first catches those before they happen, not after.
Do agents run continuously or on-demand?
No agent runs today, because agent interfaces are in development. The design varies by agent. Scout is planned for scheduled scans or on-demand triggering, and Truth Seeker for background comparison of twin state to observed state. Both depend on discovery, which is not built.
The Safety Net is designed to run per change, triggered by any proposed modification rather than on a schedule. The Architect, The Hands, and The Closer are designed to be on-demand or workflow-triggered, not continuous.
What is closed-loop automation and how do I enable it safely?
Closed-loop automation is not built. This is the design. The system would discover a required change, plan it, simulate it, get approval, execute it, and confirm success without a human touching each step. The Architect would plan, The Hands would stage, and The Closer would execute with rollback.
The safety model is part of the design: a human approves by default, so every execution plan needs a sign-off before it runs. Auto-approval would be opt-in per change type, for example low-risk DNS record additions, with manual approval kept for everything else.
The intent is a full audit trail for every execution: who approved, what changed, and the pre- and post-state.
Still have questions?
We're happy to answer anything not covered here - whether it's a technical deep-dive, a migration question, or just trying to figure out which plan fits your environment.
Usually responds within one business day.