CONNOR

CONNOR is an intelligence that builds intelligence.

It advises the simulation. It coaches the operator, and it competes against them. It does the same for an AI agent.

CONNOR

An intelligence that builds intelligence: in people, in AI agents, and in itself.

What it does

CONNOR CoachCONNOR BIAdviseCompeteGradeTradecraft IntelligenceTelemetry

Key features

  • Advises the simulation as it runs
  • CONNOR Coach helps people complete cyber tasks, working inside the task environment
  • CONNOR BI analyses your organisation's cyber task performance in 5DPRO
  • Competes against humans and AI agents alike
  • Grades what an agent actually did
  • Tradecraft built from intelligence and exercise telemetry
  • Tradecraft reached only through CONNOR
  • A RAG source behind an API and MCP, for the model you already run

In people

CONNOR coaches the operator through a mission and competes against them inside it. The coaching is specific to the mission, because the simulation was built to that mission.

In 5DPRO it takes two forms today. CONNOR Coach helps people complete cyber tasks. It can interact with the task environment and execute commands, so its help is grounded in the state of the task rather than in a description of it. CONNOR BI is a business intelligence agent that lets you analyse your organisation’s cyber task performance within 5DPRO: who completed what, how, and where the force is short.

In AI agents

The same intelligence does the same work for an AI agent: advises, coaches, competes, and grades what the agent actually did.

In itself

Behind CONNOR is Tradecraft Intelligence, a body of cyber tradecraft built from external intelligence and from the telemetry of every mission exercised through the system. Tradecraft is reached only through CONNOR.

As a RAG source

CONNOR is a cyber expert behind an API and MCP. Put it in front of the model you already run as a retrieval source, and it answers from Tradecraft Intelligence: the external intelligence and the operational intelligence generated across every previous simulation. Or put it beside the other agents as the one that checks their work.

It does not replace the model you already run. XORCISE sends each task to the right one. Tradecraft is reached only through CONNOR, so the corpus behind it is never handed over; what is handed over is the answer, grounded in it.

Talk to us about CONNOR

Tradecraft Intelligence

CONNOR's knowledge, above its intelligence.

Behind CONNOR is Tradecraft Intelligence: a body of cyber tradecraft built over a decade of scenario building, and still growing. Two things feed it: reporting and intelligence, open source and closed, and the trace of every agent run against the simulations.

A trace is not only how a mission was completed. It correlates the AI system's profile, the mission it faced and the actions it took. Processed, enriched and structured into a knowledge graph, that is the store, and CONNOR is the only way in.

Tradecraft Intelligence · filtration
Tradecraft Intelligence: two inputs filtered into one store, reached through CONNOR Open and closed source reporting and intelligence, and agent traces from runs against the simulations, fall through three stages, process, enrich and structure, into a knowledge graph, the Tradecraft Intelligence store. CONNOR answers intelligence requests from LEAP.FROG, XORCISE.CONTROL and 5D.PRO from its own knowledge and by retrieval over the store. Intelligence open and closed source Agent traces every run against the simulations AI profile what it is Mission terrain, objectives, rubrics Actions what it did reports · advisories · closed feeds Tradecraft Intelligence the store
CONNOR interprets the request, answers from what it knows and from the store LEAP.FROG simulation generation XORCISE.CONTROL XORCISE data generation 5D.PRO CONNOR Coach · CONNOR BI Two inputs. Three plates. One store. Reached only through CONNOR.

What goes in

Open and closed source reporting and intelligence, and the trace of every run against the simulations, including LEAP.FROG's own agents black-box testing them.

What a trace correlates

The AI system's profile (what it is), the mission (terrain, objectives, rubrics) and the actions (what the agent did). XORCISE profiles the model and agent, so performance reads against what produced it.

What comes out

A knowledge graph. When LEAP.FROG, XORCISE.CONTROL or 5D.PRO asks, CONNOR interprets the request and answers from what it knows and from the store. The store is never handed over.