Introducing RunaReg: a regulatory workspace, not a regulatory oracle

There is no shortage of tools that promise to “automate MDR compliance.” We built RunaReg on the opposite premise: compliance decisions belong to qualified humans, and what those humans lack is not intelligence — it’s a workspace where the regulation, the evidence, and the record of their decisions live in one connected place.

That’s RunaReg. Write it here. Cite it right. Defend it anywhere.

The architecture is the promise

RunaReg is deterministic-first, in three tiers:

1. The knowledge graph decides what applies. MDR 2017/745, its Annexes, MDCG guidance, and harmonized standards are encoded as a versioned graph with checksummed sources. The same device profile always maps to the same obligations.
2. The LLM only suggests. Drafts are generated with inline assertions, and every assertion is validated against the graph before you see it: the cited node must exist, the confidence threshold must be met, and the supporting evidence must actually be in your document. Failed assertions trigger a re-draft — not a shrug.
3. Retrieval provides reference, never decisions. Guidance and standards are surfaced for reading, not auto-applied.

A qualified human approves everything that matters, and every approval lands in an append-only, ALCOA+-oriented audit log.

What shipped this spring

The last few months turned that architecture into a working platform. Highlights from the changelog:

• Gap Analysis with clause-linked findings. Technical files are checked against MDR Annex II obligations; every finding carries its governing clause and a severity grade. Remediation is preview-then-apply, creating reviewable revisions rather than silent overwrites.
• CER Builder. A dedicated clinical evaluation workspace with an MDCG 2022-2 equivalence panel and a deterministic literature pipeline that builds its query from your equivalence data, searches PubMed, applies an evidence hierarchy filter, and writes a methodology log automatically.
• Typed placeholders in drafts. Drafted documents mark every unresolved item as a typed placeholder, split into critical items and suggestions to review — you always know what stands between a draft and a signable document.
• Evidence lineage. Every vault item records how it entered the system — manual upload, literature import, gap-analysis stub — and citation metadata (DOI/PMID) feeds a one-click Vancouver-style bibliography.
• Submission Readiness gate. Before an export, four checks run in parallel: cross-document consistency, completeness, the PRRC sign-off queue, and critical gaps. Exports are anchored by an append-only SHA-256 snapshot of the technical file manifest.
• Sentinel guidance monitoring. 40+ MDCG documents are watched in the background. A detected change writes an immutable audit row with the affected device families, and your acknowledgment of it is recorded permanently.
• Honest jurisdiction tiers. Our coverage census labels EU MDR as Full and IVDR / UKCA / Türkiye overlays as Preview — on the platform and on this website. If a coverage label ever looks generous, check the census; the website is generated from the same numbers.

Two clocks worth watching

If you operate in or sell into Türkiye, 1 February 2027 is the EU 2026/977 CE-mark cutoff that determines which devices fall under the alignment obligations due later that year. And for everyone holding legacy MDD certificates, the May 2027 end of the MDD→MDR transition window is the hard stop.

Both timelines reward starting early — gap analysis before the Notified Body’s formal clock starts is exactly the work RunaReg structures.

Founding partners

We’re onboarding a focused cohort of founding partners: manufacturers preparing Class II/III documentation, consultancies running repeatable pre-submission reviews, and PRRCs who want a sign-off trail they can defend. If that’s you, request a demo — we reply directly, within one working day.