Quantum-Inspired Optimisation. For AI agents.
Solve NP-hard combinatorial problems — vehicle routing, job scheduling, portfolio selection — that classical AI cannot handle. Runs on ordinary CPUs via 13 MCP tools. No quantum computer. No special hardware. Built into Purple8 Hyper Graph.
Clarity first
Quantum-inspired vs quantum computing
- •Requires physical qubits — fragile, expensive hardware
- •Best machines: thousands of noisy qubits in 2026
- •Limited practical business applications today
- •Cloud queue time; not real-time
- •Promising long-term, not production-ready for most use cases
- •Runs on CPU — same machine as the rest of Purple8
- •Borrows mathematical structure: QUBO, annealing, tensor networks
- •Solves VRP, scheduling, portfolio in seconds, today
- •13 MCP tools — callable by any AI agent
- •No quantum hardware, no subscriptions, no queue
quantum.* namespace
13 MCP tools. All callable by agents.
quantum.optimizeAuto-selects the best solver based on problem size and density
quantum.annealGeneric QUBO via simulated annealing with greedy descent refinement
quantum.tabuQUBO via tabu search — memory-based local search
quantum.vrpVehicle Routing Problem with capacity and time window constraints
quantum.portfolioBinary portfolio optimisation under return, risk, and sector limits
quantum.schedulingJob scheduling across machines with precedence and resource constraints
quantum.simulateTensor network (MPS) simulation and expectation value computation
quantum.vqe_stepOne variational quantum eigensolver iteration for Hamiltonian minimisation
quantum.encode_stateStore a quantum state vector in the graph, indexed by fidelity
quantum.measureSimulate measurement of a stored quantum state
quantum.compareFidelity, trace distance, and entanglement entropy between states
quantum.profileProfile solver performance on a given problem instance
quantum.job_statusCheck status of a background optimisation job
Use cases
Problems AI agents can now solve
These are NP-hard problems. Neural networks don't generalise well on them. Quantum-inspired solvers do.
Last-mile delivery routing
Route 200+ vehicles across a city with capacity limits and customer time windows. QUBO + simulated annealing finds solutions within 2–5% of optimal in seconds — on a CPU.
Portfolio construction
Select a binary portfolio of assets under expected return targets, risk tolerance, and sector concentration limits. Exact solvers can't scale; quantum-inspired can.
Manufacturing job scheduling
Assign 500 jobs to 40 machines with setup times, precedence constraints, and shift boundaries. Minimise makespan and tardiness simultaneously.
Staff rostering
Assign shifts to workers respecting qualifications, preferences, union rules, and legal constraints. QUBO encodes every constraint as a penalty term.
Space allocation (AEC)
Allocate rooms, floors, and zones to departments under adjacency, separation, and area constraints. Graph-colouring + constraint propagation via the AEC vertical.
Energy dispatch optimisation
Decide which generators to switch on/off at each interval to meet demand at minimum cost, with binary on/off decisions per asset — a natural QUBO formulation.
FAQ
Common questions
What is quantum-inspired optimisation?
Quantum-inspired optimisation borrows mathematical structures from quantum physics — QUBO formulation, simulated annealing, tensor networks — and runs them on classical CPUs. It solves combinatorial optimisation problems (VRP, job scheduling, portfolio selection) that are NP-hard and where neural networks and gradient-based methods perform poorly. No quantum computer is required.
How is this different from real quantum computing?
Real quantum computing uses physical qubits — particles manipulated via quantum mechanics. Quantum-inspired optimisation uses the mathematical structure (QUBO, annealing, tensor networks) without quantum hardware. It runs today, on ordinary CPUs, and produces good-quality solutions to NP-hard problems in seconds.
What kinds of problems can quantum-inspired optimisation solve?
NP-hard combinatorial problems: vehicle routing with capacity and time window constraints, job shop scheduling with precedence rules, binary portfolio optimisation under sector limits, staff scheduling with qualification constraints, and space allocation problems in AEC. These are problems where brute force is computationally infeasible and neural networks do not generalise well.
How do AI agents use the quantum tools in Purple8?
Via 13 MCP tools in the quantum.* namespace — quantum.vrp, quantum.portfolio, quantum.scheduling, quantum.optimize, quantum.anneal, and more. An agent describes the problem in natural language, retrieves structured data from the graph, formulates the QUBO, calls the solver, and interprets the result. The solver runs in-process, in the same Python runtime as the rest of Purple8.
Try it in 60 seconds
Developer edition is free. All quantum.* tools are included. No hardware, no cloud subscription, no special setup.