The regulatory workspace for medical device teams

Write it here.
Cite it right.
Defend it anywhere.

RunaReg maps your technical file to MDR 2017/745, surfaces compliance gaps clause-by-clause, and drafts evidence-backed remediation for your review. A qualified human makes every regulatory decision — RunaReg makes that decision defensible.

EU-hosted (Frankfurt) GAMP 5 Category 4 aligned ALCOA+ audit log
Two clocks are running
1 Feb 2027

EU 2026/977 — Türkiye CE-mark cutoff

Devices marketed in Türkiye with CE marks granted before 1 February 2027 fall under the EU 2026/977 alignment obligations, due 1 November 2027.

May 2027

MDD → MDR transition window closes

Legacy MDD certificates expire. Devices without MDR conformity cannot remain on the EU market after the transition window closes.

The connected-evidence story

One graph connects the regulation to your evidence.

Most teams keep the regulation in PDFs, the evidence in folders, and the reasoning in someone's head. RunaReg keeps all three connected: a versioned regulatory knowledge graph, an evidence vault with provenance, and an audit trail that records every decision you make.

TIER 1 — DETERMINISTIC RULES Knowledge graph decides what applies KG-anchored compliance logic · same input, same answer DECIDES TIER 2 — LLM ADVISORY Drafts and suggestions, validated against the graph Every assertion checked: node exists · confidence met · evidence present SUGGESTS TIER 3 — CONTEXTUAL REFERENCE Retrieval only — guidance and standards for reading REFERENCES human approval gate never auto-applied
Deterministic-first, in three tiers. The knowledge graph decides what applies. The LLM only suggests — and every suggestion waits for your approval. Retrieval provides reference context, never auto-applied.
REGULATION CLAUSE MDR Annex II · 1.4 node in the versioned knowledge graph SOURCE DOCUMENT Official EU text · SHA-256 checksummed before it enters the graph YOUR EVIDENCE Vault item · provenance recorded linked to the document section it supports HUMAN DECISION Sign-off · audit row · timestamp the consultant or manufacturer decides
From regulation text to a defensible decision. Every finding carries its chain: the regulation clause, the checksummed source document, the evidence in your file, and the human who approved it.
How a finding is produced

Deterministic mapping. You stay the regulatory owner.

Three checkpoints from regulation text to a reviewable finding. The same input always produces the same clause link.

Regulation text

MDR 2017/745, Annexes II / III / VIII / XIV, MDCG guidance, harmonized standards — versioned in a 1,100+ node regulatory knowledge graph with checksummed sources.

RunaReg maps your file

Each clause is linked to the matching paragraph in your technical file. Every match is reproducible and the trace is exposed for review.

You review and decide

Your PRRC or consultant approves remediation. Nothing exits the platform without a human regulatory owner.

RunaReg applies deterministic clause mapping first. Draft remediation remains review-only and never replaces a qualified regulatory owner.

Why not just use a chatbot?

Deterministic vs probabilistic — the audit difference

What changes when your output has to be reproducible at audit — not just persuasive on screen.

Capability
Generic AI tools
RunaReg
Gap review
Prompt-based answers with inconsistent structure coverage.
Deterministic review mapped to MDR structure, clauses, and evidence gaps.
Drafting confidence
Text that sounds authoritative but may fabricate study references or cite non-existent harmonized standards.
Knowledge-graph-grounded output with assertion-level checking: the cited node must exist, the confidence threshold must be met, and the evidence must be present in your document.
Traceability
No source attribution. Your PRRC has to reconstruct the reasoning chain by hand.
Every finding stamped with the graph node, reviewer, and source paragraph. Defensible at audit.
Submission control
One-click export with no reviewer gate. Easy to send something the NB will reject.
Section-by-section review with explicit sign-off before export. The human regulatory owner stays in the loop.
Jurisdiction coverage

Honest coverage tiers, straight from our knowledge graph census.

We label jurisdiction depth the way we'd want a supplier to label it for us: Full where the graph is complete, Preview where coverage is early. No padded checklists.

🇪🇺 EU MDR 2017/745

All 23 GSPRs, Annex VIII classification rules, complete Annex II/III technical-file structure, and the MDD legacy bridge.

Full
🇪🇺 EU IVDR 2017/746

Early coverage: 20 Annex I GSPR clauses and IVDR classification rules 1–3 with Class A fallback. Expanding.

Preview
🇬🇧 UK (UKCA)

Early coverage: 15 curated UK MDR 2002 conformity requirements.

Preview
🇹🇷 Türkiye (TİTCK)

Our deepest overlay in progress: 55 curated TİTCK requirements aligned with the EU 2026/977 window.

Preview

Tier labels are sourced from the same coverage census our platform reports internally.

Frequently asked

Questions teams ask before they share a file.

Does RunaReg replace the PRRC or consultant?

No. RunaReg structures the evidence, findings, and remediation workflow. A qualified human reviewer remains the regulatory owner and decision-maker.

Why do you call the mapping deterministic?

The same file and the same rule set produce the same clause link. The platform does not guess — it applies predefined matching logic and exposes the trace for review.

Can we use it before an NB quotation request?

Yes. The platform is designed for pre-submission readiness, gap identification, and remediation planning before a Notified Body starts the formal clock.

How is data hosted and protected?

Customer application data and working documents are processed and stored in the EU (Frankfurt, europe-west3), with row-level tenant isolation and an ALCOA+ oriented audit log. See the Trust & Security page for the full picture, including what we do and don't claim.

What makes the mapping deterministic — the knowledge-graph view?

Generic AI tools answer from training-time corpora. RunaReg answers from a versioned regulatory knowledge graph — MDR, MDCG guidance, and harmonized standards represented as nodes and clause links, rebuilt on every regulation update. The platform then checks each AI-drafted sentence against the graph: does the clause exist as a node? Does the cited standard? Is the evidence in your file? Same input, same graph version, same answer.

Is RunaReg an EUDAMED submission system?

No. RunaReg helps you prepare technical documentation and registration-ready data before you interact with EUDAMED. It does not submit to EUDAMED and does not replace your own submission workflow.

Request access

We are onboarding a focused set of MDR teams now.

Tell us your device class, submission stage, and current bottleneck. We reply directly to coordinate access — no automated drip, no demo theatre.

  • Manufacturers preparing Class II / III technical documentation
  • Consultancies running repeatable MDR pre-submission reviews
  • QA / RA teams building a traceable remediation workflow
We reply directly. No automated drip. Or email hello@runareg.com

Request a demo

This site currently uses no tracking cookies and no marketing pixels. Cloudflare Web Analytics is cookieless and aggregate. These settings exist so that if we ever add optional technologies, nothing loads without your opt-in.