SMRFORGE
Verifiable SMR computational engineering
smrforge.io · Open-core · Private beta
SMR design · analysis · optimization

Reactor physics your reviewers can reproduce — not just believe.

A full multi-physics toolkit for Small Modular Reactors where every result ships as a fingerprinted, independently reproducible evidence bundle — and is labeled, honestly, for how far to trust it.

pip install -e . → smrforge doctor → a screening k-eff in a minute
evidence_envelope · adr-013◆ gpg-signable
analysisgt-mhr-350
k_eff0.9051
fidelityscreening_estimate
sha-256a3f1c8…9e0b
replaybit-exact @ epoch
$ smrforge verify reviewer-kit ./bundle ✓ reproducible by a third party · exit 0
The differentiator — built for the person who doesn't trust you
01 · FINGERPRINT

Evidence envelopes

Every output is a schema-registered bundle with a SHA-256 fingerprint of its inputs and result — optionally ed25519 GPG-signed.

02 · REPLAY

Bit-exact reproduction

A pinned SOURCE_DATE_EPOCH and the replay_bundle() contract regenerate the same bytes from the same inputs.

03 · VERIFY

exit 0 = it holds

Verification is a standalone, offline check against the open Evidence Envelope schema — a third party reproduces or refutes your number in seconds, without trusting or licensing SMRFORGE.

One reproducible model, a full multi-physics stack — across LWR-SMR, HTGR (prismatic & pebble), MSR & microreactor archetypes
NeutronicsFull-core nodal, MC & transport · variance reduction
Thermal-HCoupled channel TH · conjugate heat transfer
Fuel cycleBurnup & depletion · Xe/Sm poisoning
SafetyKinetics transients · LOFC / ATWS / RIA / LOCA
UncertaintyMC / LHS / Sobol UQ · sensitivity
Nuclear dataIn-tree ENDF/B · self-shielding, TSL
GeometryHex / Cartesian / R-Z cores · power maps
EconomicsLCOE & fuel cycle · SMR factors
InteropOpenMC round-trip · Serpent, MCNP, plugins
Honest by default · open-core

What it gives you

  • A physics_fidelity label + plausibility band on every result
  • Machine-readable caveats that travel with the output
  • One-command escalation to OpenMC / Serpent for a validated k-eff

What it won't pretend

  • Screening k-eff is a trend, not a validated absolute
  • The open tier is an evaluation surface, not licensing sign-off
  • Nothing is tuned to hit a benchmark it was measured against
Communitysmrforge · MIT

Evaluation & integration — prove fit in the open.

  • Full screening stack + evidence envelopes & reviewer-kit
  • One-command archetype starter kits: smrforge new <archetype> → a runnable, baseline-checked project
  • OpenMC export/import · Serpent round-trip
  • Governance CLI, benchmarks, stable API, plugin system
Commercial
Prosmrforge_pro · FSL-1.1-MIT

Production & commercial depth for deployment.

  • Production physics kernels (deep S_N/MoC, adjoint, IQS, burnup-in-loop)
  • CAD / DAGMC · interactive 3D · advanced variance reduction
  • Confirmatory Console — re-run a submission, diff every claim, ship a signed verification report; subsidized for regulators & national labs
  • Regulatory-oriented reporting · assurance & benchmark evidence
# Screening k-eff, labeled honestly (private beta — from a checkout) import smrforge as smr r = smr.analyze_preset("gt-mhr-350") print(r["k_eff"], r["physics_fidelity"]) # 0.9051 screening_estimate
Open standard & verifier → Download PDF ↓ support@smrforge.io smrforge.io · Community MIT · Pro commercial

Export-control note: nuclear-related software may be subject to U.S. federal law (10 CFR Part 810, EAR, ITAR, OFAC) regardless of source-license terms. Community: MIT. Pro: source-available under FSL-1.1-MIT; commercial use under separate agreement. Screening results are labeled estimates, not validated or licensing-grade absolutes.