QCOS

One engine for circuits.
Four ways in.

QCOS is SoftQuantus's hybrid, hardware-agnostic engine for quantum circuit workloads: a Python SDK, a CLI, an HTTP API and an MCP server you run yourself. CPU simulators run today, exact up to 24 qubits at 0 credits; partner QPU gateways, GPU and HPC compute are in development.

24 qexact simulation, 0 credits
4SDK, CLI, API and MCP
6suitability verdicts before you run
SHA-256sealed run records
24 qexact simulation, 0 credits
4SDK, CLI, API and MCP
6suitability verdicts before you run
SHA-256sealed run records
24 qexact simulation, 0 credits
4SDK, CLI, API and MCP
6suitability verdicts before you run
SHA-256sealed run records
Four ways in

One engine.
Reach it from wherever you work.

The SDK, the CLI and the MCP server all speak the same runtime API, so a job submitted from a notebook, a terminal or an agent comes back the same way, with the same record. Keys, plans and credits are managed at platform.softquantus.com.

SDK

Python SDK

Build circuits and submit jobs from your own code and notebooks.

pip install softqcos-sdk
CLI

Command line

Run jobs, fetch records and verify them offline, from a terminal or a CI job.

pip install softqcos
API

HTTP API

The runtime contract under all of them: list backends, estimate, submit, collect artifacts.

api.softquantus.com/api/v1/runtime
MCP

MCP server

Run it yourself and any MCP-capable agent can call QCOS as a tool.

pip install softqcos-mcp
Ask first

Should this be quantum?
QCOS will tell you no.

Describe the problem and the suitability check answers before you run anything. Often the answer is classical, and that is worth knowing first.

# the same check, four ways
POST /api/v1/quantus/problems/suitability
softqcos quantus suitability
client.quantus.suitability()
qcos_quantus_suitability        # MCP tool
Six verdicts
  1. 01classical preferred
  2. 02GPU preferred
  3. 03HPC preferred
  4. 04hybrid recommended
  5. 05quantum suitable
  6. 06quantum not recommended
Backends

Simulators today.
The rest when it is real.

Simulators run on CPU, exact up to 24 qubits, at 0 credits. Partner QPU gateways, GPU and HPC compute are in development; when they open, each run will be estimated before it starts and metered through platform.softquantus.com.

CPU · 0 creditsQPU · in developmentGPU · in developmentHPC · in development

CPU simulators

Exact state-vector simulation up to 24 qubits, at 0 credits.

Partner QPUs

Gateways to partner quantum hardware are in development.

GPU and HPC

GPU simulation and HPC compute are in development.

A record with every run

Runs come back with a record sealed under a SHA-256 Merkle root, which you can verify offline.

FAQ

What people ask
before the first circuit.

SoftQuantus's hybrid, hardware-agnostic engine for quantum circuit workloads. You reach it four ways: a Python SDK, a command-line tool, an HTTP API at api.softquantus.com, and an MCP server you run yourself so any agent can call it as a tool. SynapseX Labs runs its quantum work on QCOS.

Circuits on CPU simulators, exact up to 24 qubits, at 0 credits. Partner QPU gateways, GPU and HPC compute are in development; when they open, they will be priced before you run and metered through platform.softquantus.com.

pip install softqcos-sdk for the Python SDK, pip install softqcos for the CLI (the command is softqcos), and pip install softqcos-mcp for the MCP server. Keys, plans and credits are managed at platform.softquantus.com.

Yes. The suitability check reads a description of your problem and answers with one of six verdicts: classical preferred, GPU preferred, HPC preferred, hybrid recommended, quantum suitable, or quantum not recommended. Often the answer is classical, and that is a useful answer before you spend.

The result, and a record of the run sealed under a SHA-256 Merkle root: backend, seed, parameters, results and cost. You can re-check that record offline with the CLI. A seal shows the record was not altered; it does not say who produced it.

We do not claim one for any workload. QCOS lets you test a quantum formulation on simulators for free, compare it with a classical baseline, and keep a record of exactly what ran.

Ask first.
Then run it free.

Install the SDK or the CLI, run the suitability check, and try the circuit on CPU simulators at 0 credits. Runs come back with a record you can keep.