Qiskit MCP Server

A Model Context Protocol (MCP) server that provides quantum circuit transpilation capabilities using Qiskit's pass managers. This server enables AI assistants to optimize quantum circuits for various hardware targets.

Features

• Circuit Transpilation: Transpile quantum circuits with configurable optimization levels (0-3)
• Preset Basis Gates: Support for IBM Eagle, IBM Heron, ion trap, and other basis gate sets
• Topology Support: Built-in support for linear, ring, grid, and custom coupling maps
• Circuit Analysis: Analyze circuit complexity without transpilation
• Optimization Comparison: Compare results across all optimization levels
• Dual API: Supports both async (MCP) and sync (Jupyter, scripts) usage

Prerequisites

• Python 3.10 or higher
• uv package manager (recommended) or pip

Installation

From PyPI (when published)

pip install qiskit-mcp-server

From Source

# Clone the repository
git clone https://github.com/Qiskit/mcp-servers.git
cd mcp-servers/qiskit-mcp-server

# Install with uv
uv sync

# Or with pip
pip install -e .

Quick Start

Running the MCP Server

# With uv
uv run qiskit-mcp-server

# Or directly
qiskit-mcp-server

Claude Desktop Configuration

Add to your Claude Desktop configuration (claude_desktop_config.json):

{
  "mcpServers": {
    "qiskit": {
      "command": "uv",
      "args": [
        "--directory",
        "/path/to/qiskit-mcp-servers/qiskit-mcp-server",
        "run",
        "qiskit-mcp-server"
      ]
    }
  }
}

Usage Examples

Async Usage (MCP Server / FastAPI)

from qiskit_mcp_server.transpiler import transpile_circuit

# Simple Bell state circuit (QASM2 - automatically detected)
qasm = """OPENQASM 2.0;
include "qelib1.inc";
qreg q[2];
creg c[2];
h q[0];
cx q[0], q[1];
measure q -> c;
"""

# Transpile with default settings (optimization level 2, QASM3 format)
result = await transpile_circuit(qasm)

# Transpile for IBM Heron processor
result = await transpile_circuit(
    qasm,
    optimization_level=3,
    basis_gates="ibm_heron",
    coupling_map="linear"
)

Using QPY Format

from qiskit import QuantumCircuit
from qiskit_mcp_server import dump_qpy_circuit
from qiskit_mcp_server.transpiler import transpile_circuit

# Create a circuit programmatically
qc = QuantumCircuit(2, 2)
qc.h(0)
qc.cx(0, 1)
qc.measure([0, 1], [0, 1])

# Convert to QPY (preserves exact parameters and metadata)
qpy_circuit = dump_qpy_circuit(qc)

# Transpile using QPY format
result = await transpile_circuit(qpy_circuit, circuit_format="qpy")
# Result includes transpiled circuit in QPY format (for chaining)
transpiled_qpy = result["transpiled_circuit"]["circuit_qpy"]

# Chain to another operation using QPY
result2 = await transpile_circuit(transpiled_qpy, circuit_format="qpy", optimization_level=3)

Sync Usage (Scripts, Jupyter)

from qiskit_mcp_server.transpiler import transpile_circuit, analyze_circuit

# All async functions have a .sync attribute
result = transpile_circuit.sync(qasm, optimization_level=2)

# Analyze circuit without transpiling
analysis = analyze_circuit.sync(qasm)
print(f"Circuit depth: {analysis['circuit_info']['depth']}")
print(f"Two-qubit gates: {analysis['gate_categories']['two_qubit_gates']}")

Compare Optimization Levels

from qiskit_mcp_server.transpiler import compare_optimization_levels

# Compare all optimization levels (0-3) for your circuit
comparison = compare_optimization_levels.sync(qasm)

for level in range(4):
    result = comparison['optimization_results'][f'level_{level}']
    print(f"Level {level}: depth={result['depth']}, size={result['size']}")

LangChain Integration Example:

Note: To run LangChain examples you will need to install the dependencies:

bash

pip install langchain langchain-mcp-adapters langchain-openai python-dotenv

import asyncio
import os
from langchain.agents import create_agent
from langchain_mcp_adapters.client import MultiServerMCPClient
from langchain_mcp_adapters.tools import load_mcp_tools
from langchain_openai import ChatOpenAI
from dotenv import load_dotenv

# Load environment variables (OPENAI_API_KEY, etc.)
load_dotenv()

# Sample Bell state circuit
SAMPLE_BELL = """
OPENQASM 3.0;
include "stdgates.inc";
qubit[2] q;
h q[0];
cx q[0], q[1];
"""

async def main():
    # Configure MCP client
    mcp_client = MultiServerMCPClient({
        "qiskit": {
            "transport": "stdio",
            "command": "qiskit-mcp-server",
            "args": [],
            "env": {},
        }
    })

    # Use persistent session for efficient tool calls
    async with mcp_client.session("qiskit") as session:
        tools = await load_mcp_tools(session)

        # Create agent with LLM
        llm = ChatOpenAI(model="gpt-5.2", temperature=0)
        agent = create_agent(llm, tools)

        # Run a query
        response = await agent.ainvoke(f"Transpile this circuit for IBM Heron: {SAMPLE_BELL}")
        print(response)

asyncio.run(main())

For more LLM providers (Anthropic, Google, Ollama, Watsonx) and detailed examples including Jupyter notebooks, see the examples/ directory.

API Reference

Tools

Resources

Core Functions

transpile_circuit(circuit, optimization_level=2, basis_gates=None, coupling_map=None, initial_layout=None, seed_transpiler=None, circuit_format="qasm3")

Transpile a quantum circuit using Qiskit's preset pass managers.

Parameters:

• circuit: Quantum circuit as QASM3 string, base64-encoded QPY, or QASM2 string (max 100 qubits, 10,000 gates)
• optimization_level: 0-3 (default: 2)
  • 0: No optimization, only basis gate decomposition (fastest)
  • 1: Light optimization with default layout
  • 2: Medium optimization with noise-aware layout (recommended)
  • 3: Heavy optimization for best results (can be slow for large circuits)
• basis_gates: List of gate names or preset ("ibm_default", "ibm_heron", etc.)
• coupling_map: List of edges or topology name ("linear", "ring", "grid", "full")
• initial_layout: List of physical qubit indices (length must match circuit qubits)
• seed_transpiler: Random seed for reproducibility
• circuit_format: Format of the input circuit ("qasm3" or "qpy"). Defaults to "qasm3". When "qasm3" is specified, QASM2 is also accepted as a fallback.

Returns: Dictionary with original/transpiled circuit info and optimization metrics

Note: Level 3 optimization can be very slow for circuits with >20 qubits or >500 gates. Use level 2 for faster results with good quality.

analyze_circuit(circuit, circuit_format="qasm3")

Analyze circuit structure and complexity.

Returns: Dictionary with gate counts, depth, and categorization (single/two/multi-qubit gates)

compare_optimization_levels(circuit, circuit_format="qasm3")

Compare transpilation results across all optimization levels.

Returns: Dictionary comparing depth, size, and gates for levels 0-3

Circuit Format Support

The server supports two circuit formats for input:

QPY output: All tools return circuits in QPY format (base64-encoded) for precision when chaining tools/servers.

When to use each format:

• QASM3 (input): Best for human-readable circuits and initial input
• QPY (input/output): Best for preserving exact numerical parameters when chaining tools/servers

Converting QPY to Human-Readable QASM3

To view a QPY circuit output in human-readable format, use the qpy_to_qasm3 utility:

from qiskit_mcp_server import qpy_to_qasm3
from qiskit_mcp_server.transpiler import transpile_circuit

# Transpile a circuit (returns QPY format)
result = transpile_circuit.sync(qasm_circuit, optimization_level=2)
qpy_output = result["transpiled_circuit"]["circuit_qpy"]

# Convert to human-readable QASM3
conversion = qpy_to_qasm3(qpy_output)
if conversion["status"] == "success":
    print(conversion["qasm3"])

Converting QASM3 to QPY

To convert a QASM circuit to QPY format (for full fidelity when chaining tools), use qasm3_to_qpy:

from qiskit_mcp_server import qasm3_to_qpy

qasm_circuit = '''
OPENQASM 3.0;
include "stdgates.inc";
qubit[2] q;
h q[0];
cx q[0], q[1];
'''

# Convert to QPY format
result = qasm3_to_qpy(qasm_circuit)
if result["status"] == "success":
    qpy_string = result["circuit_qpy"]
    # Use qpy_string with tools that accept QPY input

Available Basis Gate Sets

You can also provide a custom list of gate names for other hardware targets.

Available Topologies

Limits and Performance

Circuit Size Limits

To ensure reliable performance, the server enforces the following limits:

Circuits exceeding these limits will return an error with a descriptive message.

You can override these limits via environment variables:

# Allow up to 200 qubits and 50,000 gates
export QISKIT_MCP_MAX_QUBITS=200
export QISKIT_MCP_MAX_GATES=50000

Performance Recommendations

Tips:

• Use level 2 for most use cases (best balance of quality and speed)
• Use level 3 only when circuit quality is critical AND circuit is small (<20 qubits, <500 gates)
• Use level 0 or 1 for rapid prototyping and development
• The compare_optimization_levels tool helps identify the best level for your specific circuit

Transpilation Stages

The Qiskit transpiler processes circuits through six stages:

1. init: Decompose multi-qubit gates to 1 and 2-qubit operations
2. layout: Map virtual qubits to physical qubits
3. routing: Insert SWAP gates for hardware connectivity
4. translation: Convert to target basis gates
5. optimization: Reduce gate count and circuit depth
6. scheduling: Add timing and delay instructions

Testing

# Run all tests
uv run pytest tests/ -v

# Run with coverage
uv run pytest tests/ --cov=src --cov-report=term-missing

# Run specific test class
uv run pytest tests/test_transpiler.py::TestTranspileCircuit -v

Development

# Install dev dependencies
uv sync --group dev --group test

# Run linting
uv run ruff check src tests
uv run ruff format --check src tests

# Run type checking
uv run mypy src

# Run all checks
./run_tests.sh

Contributing

Contributions are welcome! Please see the CONTRIBUTING.md guide in the root of the repository.

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.