A number a reviewer can re-run.

Your physicists own the physics. SMRFORGE seals each result they produce as an evidence bundle anyone can verify offline and re-derive from scratch on the free, open engine. We don't compute your reactor; we make its numbers checkable. You don't have to trust us.

Evidence bundle

1.029633k_eff

fidelity: screening · code-verification · advisory

caseIAEA-2D · full core

published referencek = 1.029585

C−E (this mesh)+4.8 pcm

converged (Richardson)−3.0 pcm

signatureed25519 ✓

$ smrforge verifyINTEGRITY OK ✓

1 · your code

Produces the number

Your physicists run the calculation in whatever code they use. SMRFORGE does not compute your reactor, it makes the result checkable.

2 · SMRFORGE

Seals it, then signs it

Inputs, result, and provenance hashed into one manifest and signed. The bundle is self-contained, so it needs no network and no vendor to check.

3 · any reviewer

Checks it cold

$ smrforge verify confirms integrity and origin offline, importing none of our physics. $ smrforge reproduce re-derives the number on the open engine, held to the sealed band. Forge a bundle that passes and get credited.

Engagements, what we take on

Build the evidence where your reactor numbers are computed.

We wire the engine into the pipeline your team already runs, so every number comes out sealed, verifiable offline, and re-derivable from scratch by whoever reviews it.

Pipeline integrationflagship MCP and REST wiring into the pipeline your physicists already run, so sealing a result is one call and a signed, re-runnable bundle comes back per call, on every number you compute. Your physicists own the physics. We do not compute your reactor.

Independent confirmatory analysis Send one criticality or shutdown-margin number. We re-derive it on the open engine and OpenMC and hand back a signed, reproducible confirmation your reviewer can re-run.

Evidence packaging A number from your own code or tools, shipped as a sealed, verifiable, reproducible bundle for a submission.

Pre-submission V&V Code verification, a V&V matrix, and the benchmark closures a regulator reads, stood up and re-runnable.

Wire it into your pipeline →

Full detail: how an engagement works →

VerificationIAEA-2D full core closes the published reference to ≈ −3 pcm (code-verification, not measured validation). Screening-grade, honestly labeled on every result.

LineageA result can cite the sealed bundle it was built from, so a reviewer walks the whole derivation with one command (smrforge verify --chain), custody end to end, not just the final number.

EngineScreen fast in pure Python; escalate the same inputs to OpenMC, a lab-grade eigenvalue, free.

LicensingCommunity MIT (the whole engine + evidence spine, free) · Pro FSL-1.1-MIT (the signed assurance layer, coverage, watch, report, assess + an evidence console) · air-gap-safe.

Run it yourselfSource-only install, by design: git clone https://gitlab.com/smrforge/smrforge.git, then pip install -e . (Python 3.11+, standard library, no dependencies; not on PyPI, on purpose). Run a capability from the command line (smrforge run), from Python (run_capability), or through the local REST/MCP server (smrforge serve). The in-tree engine is screening-grade and runs anywhere Python does; the lab-grade OpenMC path needs OpenMC and a continuous-energy library on your host. Getting started: gitlab.com/smrforge/smrforge/-/blob/main/docs/GETTING_STARTED.md