Arvak.io Docs Try Arvak

Analyze quantum circuits with AI

Paste your QASM3, Qiskit, PennyLane, or Cirq code and Nathan will classify the problem type, find relevant papers from 1,700+ sources, and suggest optimizations for your target hardware.

Also available as arvak.nathan.analyze() in your Jupyter notebooks. Pass any Qiskit, PennyLane, or Cirq circuit directly.

Your code is automatically anonymized before analysis. Comments, variable names, string literals, and custom imports are stripped or normalized locally before leaving your machine. Only gate operations, circuit structure, and numeric values are sent for analysis. Learn more
pip install arvak[nathan] pip install arvak[qiskit] pip install arvak[pennylane] pip install arvak[cirq] pip install arvak[qrisp] pip install arvak[chemistry,notebook] pip install arvak[all]
Try an example:

Nathan AI

N

Welcome to Nathan

Paste your quantum circuit or pick a template to get started. Nathan will analyze it against 1700+ papers.

N
Hi! I'm Nathan, your quantum research assistant. Ask me about algorithms, circuit optimization, hardware selection, or anything quantum computing. I have access to 1700+ papers from arXiv, D-Wave, Qiskit, PennyLane, and Classiq.

Data Anonymization

Nathan automatically anonymizes your circuit code before it leaves your machine. This protects proprietary algorithms, researcher identities, and project context while preserving the structural information needed for analysis.

What gets removed

  • Comments (may contain names, project info, proprietary details)
  • String literals (may contain file paths, credentials, sensitive data)
  • Custom variable and register names (normalized to q0, c0, p0)
  • Custom function and class names (normalized to fn0, Cls0)
  • Non-standard imports (custom modules, internal libraries)

What gets preserved

  • Gate operations (h, cx, rz, etc.)
  • Qubit/bit declarations and indices
  • Numeric values (rotation angles, parameters)
  • Standard library imports (stdgates.inc, qiskit, cirq)
  • Circuit structure, measurement operations, and depth

Anonymization is enabled by default in both the web interface and the Python SDK. In the SDK, you can opt out with arvak.nathan.analyze(circuit, anonymize=False). Server-side anonymization is always applied as an additional layer of protection before any data reaches the LLM.