Cloud Backends

QCOS provides a unified API across all supported quantum cloud providers. This page shows how to configure each provider, authenticate, and route circuits through QCOS.

---

How QCOS connects to cloud providers

QCOS sits between your agent and the hardware. It handles authentication, circuit transpilation, calibration-aware routing, and DRR scoring — the agent only sees the QCOS API.

Agent → QCOS API → [ IBM Quantum | IonQ | AWS Braket | Azure Quantum | On-prem ]

You provide cloud credentials once in the QCOS portal or environment. From that point on, all circuits go through QCOS regardless of which hardware they execute on.

---

IBM Quantum

Supported systems: ibm-brisbane, ibm-torino, ibm-kyiv, ibm-sherbrooke, ibm-eagle (and others via IBM Quantum Network)

Authentication:

# Set in QCOS Portal or environment
export IBM_QUANTUM_TOKEN="your-ibm-quantum-api-token"
export IBM_QUANTUM_INSTANCE="ibm-q/open/main"   # or your hub/group/project

Backend IDs in QCOS:

# List IBM backends
backends = qcos.backends.list(provider="ibm")
# Returns: ibm-brisbane, ibm-torino, ibm-kyiv, ...

Notes:

• QCOS auto-transpiles circuits to IBM native gate set (ECR, SX, X, RZ)
• Calibration data is fetched every 15 minutes from IBM Quantum
• DRR scoring is available for all IBM systems

---

IonQ

Supported systems: ionq-aria-1, ionq-aria-2, ionq-forte-1, ionq-simulator

Authentication:

export IONQ_API_KEY="your-ionq-api-key"

Backend IDs in QCOS: ionq-aria-1, ionq-aria-2, ionq-forte-1, ionq-simulator

Notes:

• IonQ systems use native gate set (MS, GPi, GPi2)
• QCOS transpiles from OpenQASM 3 / Qiskit to IonQ native
• Queue depth and error rates fetched in real time via IonQ Cloud API

---

AWS Braket

Supported systems: All Braket hardware (IonQ Harmony, IQM Garnet, QuEra Aquila) and on-demand simulators

Authentication:

export AWS_ACCESS_KEY_ID="..."
export AWS_SECRET_ACCESS_KEY="..."
export AWS_DEFAULT_REGION="us-east-1"
export AWS_BRAKET_S3_BUCKET="your-braket-results-bucket"

Or use an IAM role if running inside AWS:

{
  "Action": [
    "braket:CreateQuantumTask",
    "braket:GetQuantumTask",
    "braket:CancelQuantumTask",
    "braket:SearchQuantumTasks"
  ],
  "Resource": "*"
}

Backend IDs in QCOS: aws-ionq-harmony, aws-iqm-garnet, aws-quera-aquila, aws-sv1 (simulator)

---

Azure Quantum

Supported systems: Quantinuum H-Series, IonQ (via Azure), Microsoft Resource Estimator

Authentication:

export AZURE_QUANTUM_SUBSCRIPTION_ID="..."
export AZURE_QUANTUM_RESOURCE_GROUP="..."
export AZURE_QUANTUM_WORKSPACE="..."
export AZURE_TENANT_ID="..."
export AZURE_CLIENT_ID="..."
export AZURE_CLIENT_SECRET="..."

Or use managed identity if running on Azure:

# No credentials needed — uses workload identity
export AZURE_QUANTUM_USE_MANAGED_IDENTITY=true

Backend IDs in QCOS: azure-quantinuum-h2, azure-ionq-aria, azure-microsoft-estimator

---

Google Quantum AI

Supported systems: Willow, Sycamore (via research partnerships)

Authentication:

export GOOGLE_CLOUD_PROJECT="your-project-id"
# Uses Application Default Credentials (gcloud auth application-default login)

Backend IDs in QCOS: google-willow, google-sycamore

Note: Google Quantum AI access requires a research or commercial agreement. Contact sales@softquantus.com.

---

On-premises QPUs and GPU simulators

QCOS supports on-prem deployments via the QCOS Edge Agent — a lightweight container that connects your local hardware to the QCOS runtime.

# Deploy Edge Agent
docker run -d \
  -e QCOS_EDGE_TOKEN="your-edge-token" \
  -e BACKEND_TYPE="qpu"         # or "gpu-simulator" \
  -e BACKEND_ENDPOINT="http://localhost:5000" \
  -p 8080:8080 \
  softquantus/qcos-edge-agent:latest

Once registered, your on-prem backend appears in qcos_list_backends with ID format edge-{your-name}.

---

GPU simulators (LUMI / EuroHPC)

QCOS integrates natively with the LUMI supercomputer GPU partition and other EuroHPC systems for large-scale circuit simulation.

# Route to LUMI GPU simulator — up to 34 qubits, statevector
result = qcos.jobs.run(
    circuit=circuit,
    backend="lumi-gpu-sv",
    shots=100000,
)

Available simulator backends:

ID	Description	Max qubits	Provider
lumi-gpu-sv	Statevector on NVIDIA A100	34	EuroHPC LUMI
lumi-gpu-dm	Density matrix simulation	20	EuroHPC LUMI
aws-sv1	SV1 statevector	34	AWS Braket
ibm-aer-sv	Aer statevector	32	IBM Quantum (cloud)
local-aer	Aer on your machine	30	Local (Edge Agent)

---

Backend selection in agents

An agent can select a backend based on live criteria:

# Ask the agent: pick the best backend for a 20-qubit circuit
response = client.messages.create(
    model="claude-sonnet-4-6",
    mcp_servers=[{ "type": "url", "url": "https://mcp.softquantus.com/sse", ... }],
    messages=[{
        "role": "user",
        "content": "I need to run a 20-qubit circuit with 50,000 shots. Pick the backend with the best DRR score and lowest queue depth right now."
    }],
)
# Agent calls qcos_list_backends, filters, selects, then submits

---

Next steps

• Agent Integration Guide — multi-step agent workflows
• MCP Tools Reference — qcos_list_backends, qcos_submit_circuit
• QCOS Pricing — cost per backend per shot