MCP AI Integration¶

Tested with: Quanta SDK v0.8.1

What You'll Learn¶

Connect Quanta to Claude, GPT, and other AI assistants using the Model Context Protocol (MCP).

Prerequisites¶

01 — Getting Started
Claude Desktop or a MCP-compatible AI client

What is MCP?¶

MCP (Model Context Protocol) lets AI assistants call tools directly. Quanta exposes 14 quantum tools that any MCP client can use:

Tool	Description
`run_circuit`	Build and run any circuit
`create_bell_state`	One-click Bell state
`grover_search`	Quantum search
`shor_factor`	Factor integers
`simulate_noise`	Noisy simulation
`list_gates`	Show all 31 gates
`explain_result`	AI-friendly result explanation
`monte_carlo_price`	Option pricing
`qaoa_optimize`	Combinatorial optimization
`cluster_data`	Quantum clustering
`run_on_ibm`	Submit to IBM hardware
`ibm_backends`	List IBM backends
`ibm_job_result`	Poll IBM job
`draw_circuit`	ASCII circuit diagram

Setup — Claude Desktop¶

Install the MCP server locally:

# Install Quanta and fastmcp
pip install quanta-sdk fastmcp

# Register with Claude Desktop
fastmcp install quanta/mcp_server.py --name "Quanta Quantum SDK"

Then in Claude Desktop, you can say:

"Create a Bell state and explain the results"

Claude will call create_bell_state → explain_result automatically.

Setup — Cloud Deployment¶

For always-on MCP access:

# Run as SSE server (Cloud Run, Lambda)
python -m quanta.mcp_server --transport sse --port 8080

Example AI Conversations¶

"Factor the number 15"¶

→ AI calls shor_factor(N=15) → Returns {factors: [3, 5]}

"Search for value 7 in a 4-qubit space"¶

→ AI calls grover_search(target=7, num_bits=4) → Returns results

"Price a European call option with spot=100, strike=105"¶

→ AI calls monte_carlo_price(...) → Returns quantum vs classical prices

"Show me the noise effect on a Bell state at 5% error"¶

→ AI calls simulate_noise(circuit="bell", error_rate=0.05) → Returns fidelity

SDK Info via MCP¶

# The MCP server exposes SDK metadata
import json
sdk_info = {
    "name": "quanta-sdk",
    "version": "0.8.1",
    "tools": 14,
    "gates": 25,
    "algorithms": 12,
}
print(json.dumps(sdk_info, indent=2))

Try It Yourself¶

Install the MCP server in Claude Desktop and ask it to explain quantum entanglement with a live Bell state
Ask Claude to compare noisy vs ideal simulation
Try "Factor 21 using Shor's algorithm" in conversation

Learning Path Complete! 🎉¶

You've completed all 8 tutorials:

Tutorial	Topic
01	Getting Started — first circuit
02	Gates and Circuits — all 31 gates
03	Simulation — noise and fidelity
04	Algorithms — Grover, QAOA, VQE, Shor
05	IBM Hardware — real quantum computers
06	QML — quantum machine learning
07	QEC — error correction
08	MCP — AI integration

Next steps: - Cookbook — Copy-paste recipes for common tasks - Migration Guide — Coming from Qiskit? - GitHub — Contribute!