NwTwin - your Survivability Digital Twin

The problem

Your network survives every failure, until one it never rehearsed

Redundancy on a diagram is not resilience in an outage. A single tandem node can quietly carry a third of your point-to-point routes, and you find out during the incident. NwTwin replaces the guesswork with a number: build a replica of the network, fail things in it on purpose, and measure what actually stays reachable.

The twin

Draw the network, or import it from your inventory: core/tandem nodes, access nodes, links and trails. It is a working replica, not a picture, so every change can be scored before anyone touches production.

The index

The Network Survivability Index (NSI) runs the topology through failure scenarios (one node down, one link down, two links down, node + link) and scores connectivity and capacity network-wide, and per node and per link.

The analysis

Scores become advice in plain language: the link worth adding and by how much it helps, the link you can retire for free, the fastest recovery from the current outage. Ask follow-up questions in the app, or from an AI agent over MCP.

How it works

Model it, break it, rank it

One loop, four steps. The live network is never in the blast radius.

Model the network

Draw it in the browser or import the topology as JSON. Nodes are typed core/tandem or access; links and trails carry the routes. This is the twin, and it lives entirely on your side.

Fail things on purpose

Mark any node or link as down and run the scenario set: Base, B+1N (one node lost), B+NL (node and link), B+1L and B+2L (one and two links lost). Simulation runs in a background worker, so a large network keeps the UI responsive.

Read the score, not the diagram

Every scenario returns a survivability score network-wide and for each node and link: how many point-to-point routes survive, and how much capacity with them. The weak points stop being a matter of opinion.

Act on the advice

NwTwin names the link worth adding and the percentage it buys you, the links you can retire without losing a point, and the fastest recovery from the outage you are in. Ask it anything else in the AI wizard, or let an agent do it over MCP.

Key benefits & use cases

Six ways teams put the twin to work

Planning, operations, and the incident call: the same index answers all three.

Plan for survivability

Simulate the network under different loads and failure conditions: cyberattacks, hardware failures, equipment and software malfunctions, extreme climate events, configuration mistakes. A sandbox for the scenarios you cannot rehearse for real.

Recovery prioritisation

When several things are down at once, the twin evaluates the alternatives and ranks them: restore this link first, because it returns the most survivability per hour of work. Fewer debates, shorter outages.

Operations management

Score planned maintenance before the window opens. See what a truck roll, a card swap or a node reboot does to tolerance while it is out, and pick the window where the network can actually afford it.

Comprehensive AI analysis

Detailed reports on network performance with actionable advice: bottlenecks, forward-looking risks of failure, service availability exposure, and inventory accuracy gaps the topology itself gives away.

Customisable

The algorithms are configurable to the way your network is actually built and run, across telecommunications, finance, healthcare and government, not one fixed model of what a network is supposed to look like.

Open API & MCP

An open HTTP API and a Model Context Protocol server put the twin next to your service assurance and fulfilment systems, and inside your AI tooling, for continuous monitoring rather than a quarterly study.

AI & MCP

"Claude, which link should we restore first?"

The twin is not just a UI. Point an AI client at it and every network becomes something you can interrogate in plain language.

Connect an AI client

Start the app, then add one MCP entry pointing at http://localhost:3000/mcp.

{
  "mcpServers": {
    "nwtwin": {
      "url": "http://localhost:3000/mcp"
    }
  }
}

Then just ask

Seven tools cover the whole surface: list, inspect, edit, fail, simulate, read results, analyse:

"Which networks are scoring below 0.95?"
"Fail TANDEM-3 and the Ring-2 link, then re-run the simulation."
"What single link would improve Metro Core the most?"
"Write the survivability summary for this week's change board."

All MCP tools