Runner-agnostic execution
Run your existing suites where they live — GitHub Actions today, runner-agnostic by design. Sentula orchestrates; you keep your pipeline.
For FDA medical-device & SaMD teams
Compliance evidence,
Compliance evidence,
generated — not assembled.
Sentula runs your medical device's own test suites and turns every run into traceable, IEC 62304–aligned FDA evidence — ready for 21 CFR Part 11 sign-off in the QMS you already run.
- Runner-agnostic
- Deterministic — replay any line of the record
- Under your QMS, over your pipeline
Purpose-built for FDA-regulated medical software — from Class II devices to AI/ML-enabled Software as a Medical Device. Fluent in the standards your auditors already expect:
- IEC 62304
- ·
- ISO 13485
- ·
- ISO 14971
- ·
- 21 CFR 820 (QMSR)
- ·
- 21 CFR Part 11
The problem
Validation resets to zero
Validation resets to zero
on every change.
Today your team rebuilds compliance evidence by hand after every release — the manual binder that can consume a large share of a device's cost to market. Then the next commit resets it.
~0%
of the cost to bring a medical device to market is regulatory & compliance work. (Stanford, 2010)
6 → 1,200+
FDA-authorized AI/ML devices, 2015 to today — the cadence of change is exploding.
Every release
re-proves the same requirements, by hand, from zero. The change rate is only climbing.
How it works
Run the tests. Get the evidence. Sign it.
Sentula makes compliance evidence a by-product of running your tests — not a project you start after the release.
-
01
Run
Execute your existing suites where they already run — GitHub Actions today, runner-agnostic by design. No rip-and-replace.
-
02
Capture
Collect the proof each run produces — logs, artifacts, and visual proof of the software behaving as specified — every artifact hashed and attributed the moment it lands.
-
03
Trace
Link every result back to the requirement it satisfies — requirement → risk → test → evidence, kept live.
-
04
Sign
Route to QA for 21 CFR Part 11 sign-off with a tamper-evident audit trail — then deliver a signed, track-changes-ready record into the QMS you already run.
What's missing
The half your backbone
The half your backbone
leaves to you.
Traceability backbones store the evidence you upload and run tests through external integrations. Sentula adds the missing half: a built-in executor that generates the evidence, then assembles it into the sign-off-ready, track-changes document your QMS expects — feeding whatever backbone you already run.
Backbone-only tools
- — Store evidence you attach by hand
- — Run tests through external integrations
- — No built-in, runner-agnostic executor
- — No autonomous computer-use capture
Sentula
- ✓ Built-in, runner-agnostic test executor
- ✓ Evidence generated from every run — with autonomous computer-use capture as the differentiating edge we're building
- ✓ Live requirement → risk → test → evidence trace
- ✓ 21 CFR Part 11 sign-off — feeding the QMS you already run, not replacing it
What Sentula is
- ✓ The execution-and-evidence layer that runs your SDLC's own tests and turns every run into deterministic, sign-off-ready proof.
- ✓ The engine that feeds a controlled, track-changes-ready record into the QMS/ALM backbone you already run.
What Sentula is not
- — Another eQMS or system of record — your QMS stays; we feed it.
- — A competitor to your traceability backbone — we own the runner/evidence seam it delegates, not the backbone.
What you get
An execution-and-evidence layer
An execution-and-evidence layer
over the stack you already run.
Under your QMS, over your test pipeline.
Evidence capture, every run
Test evidence is captured on every run — logs, artifacts, and visual proof of the software doing exactly what the spec requires — hashed, attributed, and linked as it lands.
Live traceability graph
Every result is linked requirement → risk → test → evidence — the IEC 62304 / ISO 14971 trace kept live instead of going stale between audits. An agentic trace-keeper proposes missing links and flags coverage holes; every proposal lands as an attributed change a human approves.
21 CFR Part 11 sign-off
Your QA team holds approval and signs each record with a tamper-evident audit trail. Authority stays with people, not the machine.
Regenerates on every change
When code or a model changes, the evidence regenerates rather than resetting your validation to zero.
Feeds the QMS you already run
Delivered as a signed evidence bundle you upload — or an API push where that's cleaner. Whatever QMS you run, or none: no lock-in to one backbone, no rip-and-replace.
Why now
The rules just turned toward
The rules just turned toward
test-generated digital evidence.
Three regulatory shifts now favor risk-based, executable proof over hand-assembled binders — exactly what Sentula produces.
In force · Feb 2, 2026
FDA QMSR
The Quality Management System Regulation incorporates ISO 13485:2016 by reference — aligning U.S. device quality with a globally harmonized standard.
Final · Feb 3, 2026 (rev.)
CSA guidance
Computer Software Assurance for Production and Quality Management System Software — the Feb 3, 2026 revision supersedes the 2025 final. Risk-based, test-evidence-driven assurance is unchanged: less paperwork, more proof from the tests themselves.
Final · Dec 4, 2024
PCCP guidance
Predetermined Change Control Plans let AI-enabled devices evolve under a pre-authorized envelope — change becomes continuous, and so must the evidence.
The trust model
Agents generate.
Agents generate.
Math seals.
People sign.
Sentula is agentic where that saves you weeks — proposing trace links, drafting the record, capturing visual proof of the running software. And it is deterministic where an auditor needs it: pass/fail comes only from executing your tests, every artifact is hashed into a tamper-evident chain, and the same inputs always regenerate the same record. Every AI contribution arrives as an attributed proposal a human approves — and the permission model itself makes it impossible for an agent to sign. Not a policy. An invariant.
Who it's for
Built to earn a yes
Built to earn a yes
from every seat.
A tool only enters a regulated stack with a head-nod from engineering, product, quality, and the executive team. Sentula is designed so each seat keeps what it owns — and sheds the manual work.
Keep your pipeline
Your CI and runners stay exactly as they are. Sentula instruments the suites you already run and turns each run into attributed, traceable evidence — the evidence-to-document grunt work stops landing on engineering.
Ship without the compliance gate
When code or a model changes, the evidence regenerates instead of resetting validation to zero — compliance stops being the step that adds months to every release.
Proof you can stake an audit on
Deterministic and reproducible: the same inputs yield the same proof, your team holds every Part 11 signature, and you can replay any line of the record in front of an auditor.
Your QMS stays in charge
Sentula hands your QMS a controlled, sign-off-ready record — a signed evidence bundle you upload, or an API push where that's cleaner. Nothing about your Part 11 process changes.
Add a layer, not a migration
An evidence layer over the pipeline you already run — no re-architecting the SDLC, no rip-and-replace. Pass/fail comes from your tests, not a model's judgment.
Attack the biggest cost line
Regulatory and compliance work is the biggest share of a device's cost to market — and it recurs on every release. Sentula turns that recurring manual cost into a by-product of the tests you already run.
Stop assembling proof.
Start generating it.
Join the early-access cohort. We're onboarding a small set of FDA-regulated medical-device and SaMD teams ahead of general availability.