SynapseX Labs

AI for quantum
research. With receipts.

A research agent with its own laboratory: it drafts the experiment, probes it on free simulators first, escalates to real hardware only with your approval — and freezes every run into a reproducible record.

Open the lab — freeRead the docs
Labs live on every accountFree simulators liveReal QPU early accessGPU waitlist
$0quantum simulators, every plan
100free compute runs each month
24 qexact statevector simulation
+18.5 ptsrouting fidelity vs naive, on real quantum hardware
$0quantum simulators, every plan
100free compute runs each month
24 qexact statevector simulation
+18.5 ptsrouting fidelity vs naive, on real quantum hardware
$0quantum simulators, every plan
100free compute runs each month
24 qexact statevector simulation
+18.5 ptsrouting fidelity vs naive, on real quantum hardware
The research loop

From question
to receipt.

I
step 1 / 4running

Brief the agent

Describe the question in plain language. The agent drafts hypotheses, circuits and analysis code in a persistent Labs workspace — Jupyter notebooks included.

Use cases

What researchers run here.

Inside the quantum lab

Algorithm bring-up

Prototype and debug quantum algorithms against exact simulation, sweeping parameters from Jupyter notebooks the agent reads and edits with you — and catch the bug before it costs real shots.

Noise & mitigation studies

Characterize noise models and test zero-noise extrapolation before hardware: our ZNE engine measures 6.5× observable-error reduction in noisy simulation, at 4× shot cost.

Hardware selection

Browse QEC code families and a live catalog of 33 providers through public, key-free APIs — then let the agent argue the trade-offs for your device.

Chemistry & materials

Ground-state and small-ansatz studies on the free simulator; escalate only the survivors to metered hardware.

Reproducible papers

Run manifests and evidence bundles designed to be checked by someone who wasn't in the room.

Benchmarks

Numbers
with receipts.

Quantum marketing runs on multipliers. Ours come with job IDs. QuantumOS measures every backend the same way, whoever makes it — and every number we publish is labeled simulation or hardware, negative results included.

156 q
largest QPU we validate on
+18.5
fidelity points, routing vs naive
2,181
tests in the quantum core
Read the docs
evidence ledgerOperational
Real-hardware validation
4 circuits × 4,096 shots on a 156-qubit QPU — provider job IDs published
hardware
Predictive shot routing
+18.5 fidelity points vs naive placement · rank correlation 0.855
hardware
ZNE error mitigation
6.5× observable-error reduction at 4× shot overhead
simulation
Evidence verifier
median 0.0115 ms per bundle · single-byte tampers detected
benchmark
Why SynapseX

Built for scientists
who check.

Trust in quantum results shouldn't require trusting us. The workflow assumes you'll verify — so every layer leaves something verifiable behind.

sha256 manifests$0 simulatorsapproval before spendsim vs hardware, labeled

Reproducible by construction

Code, dependencies and backend are hashed into a manifest at dispatch — the run is re-runnable, the record tamper-evident.

Cheapest probe first

Free simulation answers most questions. Paid hardware asks for an estimate and your approval before a credit moves.

Public where it can be

QEC code families and the provider catalog are open endpoints — no API key to read them. Four SDKs on PyPI.

Honest results policy

Simulation and hardware numbers are never mixed. When a method loses — like ZNE on shallow circuits — we publish that too.

Toolchain

Works where
you already work.

Install from PyPI, keep your notebooks, script the same lab the agent uses.

synapsexPython SDK · PyPI
softqcos-sdkquantum SDK · PyPI
synapsex-platformplatform SDK · PyPI
softqlibalgorithm library · PyPI
Jupyternotebooks, native
MCPagent protocol
Python3.10+ runtime
SynapseX CLIterminal
SynapseX DesktopmacOS · Linux · Windows
synapsexPython SDK · PyPI
softqcos-sdkquantum SDK · PyPI
synapsex-platformplatform SDK · PyPI
softqlibalgorithm library · PyPI
Jupyternotebooks, native
MCPagent protocol
Python3.10+ runtime
SynapseX CLIterminal
SynapseX DesktopmacOS · Linux · Windows
synapsexPython SDK · PyPI
softqcos-sdkquantum SDK · PyPI
synapsex-platformplatform SDK · PyPI
softqlibalgorithm library · PyPI
Jupyternotebooks, native
MCPagent protocol
Python3.10+ runtime
SynapseX CLIterminal
SynapseX DesktopmacOS · Linux · Windows
SynapseX DesktopmacOS · Linux · Windows
SynapseX CLIterminal
Python3.10+ runtime
MCPagent protocol
Jupyternotebooks, native
softqlibalgorithm library · PyPI
synapsex-platformplatform SDK · PyPI
softqcos-sdkquantum SDK · PyPI
synapsexPython SDK · PyPI
SynapseX DesktopmacOS · Linux · Windows
SynapseX CLIterminal
Python3.10+ runtime
MCPagent protocol
Jupyternotebooks, native
softqlibalgorithm library · PyPI
synapsex-platformplatform SDK · PyPI
softqcos-sdkquantum SDK · PyPI
synapsexPython SDK · PyPI
SynapseX DesktopmacOS · Linux · Windows
SynapseX CLIterminal
Python3.10+ runtime
MCPagent protocol
Jupyternotebooks, native
softqlibalgorithm library · PyPI
synapsex-platformplatform SDK · PyPI
softqcos-sdkquantum SDK · PyPI
synapsexPython SDK · PyPI
FAQ

Questions researchers ask
before the first run.

It drafts them. The agent turns a question in your own words into candidate circuits, parameter sweeps and analysis code inside a persistent workspace, then runs the cheapest probe that can tell the candidates apart. You review the reasoning and the code — nothing runs on paid hardware without your approval.

Nothing. Exact statevector simulation up to 24 qubits runs at 0 credits on every plan, plus 100 free compute runs each month. That is the product, not a trial — most research questions are answered and most dead ends discarded before anything is billable.

No. Describe the problem in your field's language — chemistry, materials, optimization, machine learning. The agent decides what is classical, what is quantum, and what is not worth running at all, and explains the trade-off before spending anything.

Every dispatch freezes a manifest: a sha256 hash of each input file, the dependency set, the entrypoint, the execution mode and the backend identifier, hashed over canonical JSON. The record travels with the result, so a reviewer who was not in the room can re-run the same job and detect if anything was altered.

We are hardware-agnostic by design. QuantumOS presents every backend — our simulator fleet today, partner QPUs in early access — through one interface, measures them all the same way, and reports quality from our own instrumentation rather than a vendor data sheet. The catalog lists only what is actually live.

An SDK gives you a circuit library. This gives you the laboratory around it: an agent that plans the experiment, a metered runtime that picks the cheapest backend able to answer, credits that reserve before a run and settle for actual usage, and a reproducible record at the end. Our Python SDKs are published on PyPI if you would rather script it yourself.

The run parks with an estimate and waits for you. Nothing is reserved until you approve it; on approval, credits reserve before execution and settle for what was actually consumed, with the remainder released. Real-QPU execution is in early access — GPU and HPC tiers are on the waitlist and labeled as such everywhere on the platform.

Yes. The runtime is MCP-native, so any MCP-capable agent can list backends, request an estimate, submit a job and fetch the resulting record. There is also a headless HTTP API and a CLI. Our models are the best tenants of the lab, not the only ones.

Put an AI scientist
in your lab.

Start free: simulators cost nothing and your first 100 runs are included every month. Escalate to real hardware only when the evidence says so.

Open the lab — freeGet QPU credits

SynapseX Labs · pip install synapsex · chat.synapsex.ai