Claude DesktopでQiskitおよびQiskit Aerを使用して量子回路を作成、可視化、シミュレーションできるModel Context Protocol (MCP) サーバーです。
Claude Desktopに接続すると、以下のような自然な会話が可能になります。
「ベル状態を作成し、1024ショットでシミュレーションして」
「GHZ回路の図を見せて」
「ベル状態を高いノイズ設定で実行し、何が変わったか説明して」
「このQASM文字列からカスタム回路を作成して」
Claudeは自動的に基盤となるツールを呼び出すため、手動でツールを起動する必要はありません。
Related MCP server: Qiskit MCP Server
3つのツール: create_circuit, visualize_circuit, run_simulation
5つのプリセット回路: Bell状態, GHZ-3, 重ね合わせ, ドイッチュ・ジョサ, ランダム-4
カスタムQASM: 有効なOpenQASM 2.0文字列をすべて受け入れ可能
ノイズモデリング: 理想的、低ノイズ (p=0.001)、高ノイズ (p=0.01) を脱分極誤差でサポート
1つのリソース: resource://noise-presets — 利用可能なノイズ設定をリスト化
1つのプロンプト: simulate_walkthrough — ガイド付き会話テンプレート
Python 3.13+
Anaconda または任意のPython環境マネージャー
git clone <repo-url>
cd qiskit-sim-mcppip install -e .これにより、mcp[cli]、qiskit、qiskit-aerがPython環境にインストールされます。
python -c "
import asyncio
from quantum_mcp_demo.server import mcp
tools = asyncio.run(mcp.list_tools())
resources = asyncio.run(mcp.list_resources())
prompts = asyncio.run(mcp.list_prompts())
print('Tools:', [t.name for t in tools])
print('Resources:', [r.uri for r in resources])
print('Prompts:', [p.name for p in prompts])
"期待される出力:
Tools: ['create_circuit', 'visualize_circuit', 'run_simulation']
Resources: [AnyUrl('resource://noise-presets')]
Prompts: ['simulate_walkthrough']uv run mcp dev src/quantum_mcp_demo/server.pyブラウザで http://localhost:6274 を開きます。Claude Desktopに接続する前に、Toolsタブを使用して各ツールを手動で呼び出してみてください。
Claude Desktopの設定ファイルを見つけます:
OS | パス |
macOS |
|
Windows |
|
mcpServersオブジェクト内に以下のエントリを追加します:
{
"mcpServers": {
"qiskit-sim-mcp": {
"command": "/opt/anaconda3/bin/python",
"args": [
"-m",
"quantum_mcp_demo.server"
]
}
}
}注意:
/opt/anaconda3/bin/pythonは実際のPythonパスに置き換えてください。which pythonを実行して確認できます。
Claude Desktopを完全に終了し（macOSではCmd+Q）、再度開きます。新しい会話で、チャット入力欄のハンマーアイコン（🔨）が表示されていれば、MCPツールが読み込まれています。
「ベル状態の回路を作成し、図を表示して、理想的なノイズと高いノイズで1024回シミュレーションして。」
Claudeは以下を呼び出します:
create_circuit(preset="bell") → circuit_idを取得
visualize_circuit(circuit_id=...) → H + CNOTゲートの図を表示
run_simulation(circuit_id=..., shots=1024, noise_preset="ideal") → 約50% |00⟩, 約50% |11⟩
run_simulation(circuit_id=..., shots=1024, noise_preset="high_noise") → カウントが劣化し、|01⟩や|10⟩が出現
「この回路を実行して: OPENQASM 2.0; include "qelib1.inc"; qreg q[2]; creg c[2]; h q[0]; cx q[0],q[1]; measure q -> c;」
ClaudeはQASM文字列を解析し、同じシミュレーションパイプラインを通じて実行します。
「理想的な状態と高いノイズ環境下でのGHZ状態を比較して。ノイズは量子もつれにどのような影響を与えますか？」
src/quantum_mcp_demo/
├── server.py # MCP server entry point
├── circuits/
│ ├── presets.py # 5 named preset circuits
│ └── store.py # In-memory circuit storage (UUID-keyed)
├── tools/
│ ├── create_circuit.py # Tool: create from preset or QASM
│ ├── visualize.py # Tool: HTML diagram + gate JSON
│ └── simulate.py # Tool: AerSimulator + noise modeling
├── resources/
│ └── noise_presets.py # Resource: noise configuration list
├── prompts/
│ └── walkthrough.py # Prompt: guided simulation conversation
└── utils/
├── serialization.py # numpy → Python int conversion
└── errors.py # MCP error response builderプリセット | 量子ビット | デモンストレーション | ||
| 2 | 量子もつれ — ` | 00⟩ | 11⟩`で50/50に分割 |
| 3 | 多量子ビットの量子もつれ | ||
| 1 | アダマールゲートによる等しい重ね合わせ | ||
| 3 | 量子アルゴリズム — 干渉 | ||
| 4 | ノイズストレステスト — 固定シードのランダムゲート |
プリセット | 脱分極 p | 説明 |
| なし | 完璧なシミュレーション |
| 0.001 | 現実的な近未来デバイス |
| 0.01 | 強力なノイズ — 目に見える劣化 |
Claude Desktopにハンマーアイコンが表示されない
設定ファイルを編集した後、Claude Desktopを完全に終了して再起動してください
設定JSONが有効であることを確認してください（末尾のカンマなどがないか）
Pythonパスを確認してください: which python
python -m quantum_mcp_demo.server が失敗する
プロジェクトルートから pip install -e . を実行したか確認してください
Pythonのバージョンを確認してください: python --version (3.13以上が必要です)
MCP Inspector実行時にポート競合が発生する
kill -9 $(lsof -t -i :6274) 2>/dev/null; kill -9 $(lsof -t -i :6277) 2>/dev/nullシミュレーションが予期しないカウントを返す
ショット数の不一致は自動的に検出され、isError: trueとして返されます
エラーが発生する場合は、shotsを減らしてみてください（例: 512）
Create a quantum circuit from a named preset or an OpenQASM 2.0 string. Provide exactly one of: preset (bell|ghz_3|superposition|deutsch_jozsa|random_4) or qasm.
| Name | Required | Description | Default |
|---|---|---|---|
| preset | No | ||
| qasm | No |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description must carry full burden. It only states 'create', which implies mutation, but does not disclose side effects, permissions, rate limits, or whether the operation is reversible. Minimal behavioral context beyond the action.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is two sentences, front-loaded with purpose, and contains no unnecessary words. Every sentence adds value, making it highly concise.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given that an output schema exists (implied), the description need not explain return values. It covers the essential parameter modes and preset list. However, it lacks error handling guidance (e.g., what happens if both or neither are provided) and interaction with sibling tools, but the description remains adequate for a simple creation tool.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The schema has 0% description coverage, so the description compensates by specifying that exactly one parameter must be provided and listing the preset options. It clarifies that qasm is an OpenQASM 2.0 string, adding semantic value. However, it does not detail the format or constraints further.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool creates a quantum circuit from a named preset or an OpenQASM 2.0 string, listing specific preset options. This distinguishes it from siblings (run_simulation, visualize_circuit) which perform different actions.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage for circuit creation but does not explicitly state when to use this tool over alternatives or provide exclusions. It lacks guidance on prerequisites or context for using presets versus QASM.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Run a stored circuit on AerSimulator with optional depolarizing noise. noise_preset: ideal | low_noise | high_noise. Returns counts, probabilities, metadata.
| Name | Required | Description | Default |
|---|---|---|---|
| circuit_id | Yes | ||
| shots | Yes | ||
| noise_preset | Yes |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden. It mentions returns (counts, probabilities, metadata) but does not state whether the operation is read-only, has side effects, or requires specific permissions. The behavior is partially transparent.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is two sentences, front-loads the core purpose, and presents noise options succinctly. No wasted words.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a tool with 3 required parameters, no annotations, and an output schema, the description covers the platform and noise but omits parameter constraints and fails to distinguish output structure beyond mentioning three fields. It is adequate but not rich.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
With 0% schema description coverage, the description must compensate. It explains noise_preset with enumerated values, but leaves circuit_id and shots undocumented (no hint on circuit_id source or shots default/range). Only 1 of 3 parameters gains meaning.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool runs a stored circuit on AerSimulator with optional noise, and the verb 'Run' paired with 'stored circuit' distinguishes it from sibling tools 'create_circuit' and 'visualize_circuit'.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage context (run a stored circuit) and provides specific noise preset options (ideal, low_noise, high_noise), but does not explicitly exclude alternatives or state prerequisites like needing a pre-existing circuit.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Visualize a stored circuit as HTML and a structured gate list for narration.
| Name | Required | Description | Default |
|---|---|---|---|
| circuit_id | Yes |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, and the description does not disclose behavioral traits such as whether the operation is read-only, any side effects, or prerequisites (e.g., circuit must exist). This leaves uncertainty about its safety and failure modes.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, efficient sentence with no unnecessary words. It front-loads the key action and outputs.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the presence of an output schema and no annotations, the description is adequate but incomplete. It fails to specify error conditions, the format of the gate list, or prerequisite existence checks for the circuit.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The single parameter 'circuit_id' has 0% schema description coverage, and the description does not add any meaning beyond the schema's title. No guidance on what constitutes a valid circuit_id or how to obtain it.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the verb 'visualize' and resource 'stored circuit', and specifies the output as HTML and a gate list for narration. This distinguishes it from siblings 'create_circuit' and 'run_simulation'.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No explicit when-to-use or when-not-to-use guidance is provided. Usage is implied by the context of having a stored circuit to visualize, but alternatives like run_simulation are not differentiated.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v0.1.0create_circuitrun_simulationvisualize_circuitScored across 3 tools
Each tool addresses a distinct phase of quantum circuit workflow: creation, simulation, and visualization. No overlap in functionality.
All tools follow a consistent verb_noun convention (create_circuit, run_simulation, visualize_circuit), making their purpose immediately clear.
Three tools cover the core lifecycle adequately. While more tools (e.g., list, delete) could be added, the current count is reasonable for a focused server.
The tools enable creating, running with noise, and visualizing circuits. Missing operations like listing or deleting circuits are minor gaps but do not severely hinder basic usage.
MCP server for OpenAI API (chat completions, image generation, embeddings) via AceDataCloud
Augments MCP Server - A comprehensive framework documentation provider for Claude Code
Use AI models for chat, image, and video generation from Claude Code and other MCP hosts.
MCP server for Qwen Image 3 AI image generation
