Limited access

Structured production of EU MDR technical documentation

End-to-end generation of Annex II/III technical documentation — evidence-grounded, section-by-section, with every claim classified and traceable to its regulatory source.

Workflow

From empty workspace to finalized documentation

Each workspace is an isolated, per-product environment — evidence, drafts, and approvals never cross between products.

Define Your Product
Open a new workspace and enter the device profile: name, class, intended purpose, target users, patient group, and device characteristics (sterile, software, implant, biocompatibility requirements).
Upload Evidence
Upload existing reference documents — test reports, risk analyses, clinical data, manufacturing records. Each document is tagged with its TD category and evidence type, then automatically parsed and semantically indexed.
Select a Section & Generate
Choose from 54 TD templates. The system runs a 5-step deterministic pipeline: it loads the template structure, reads device identity from the product memory, binds uploaded evidence through a three-layer semantic matching strategy, compiles the regulatory context with cross-section consistency labels, and generates a structured draft. Four of these steps involve no AI — the language model is invoked only for the final text generation.
Edit & Approve
Review the draft, edit inline, then approve. Approval is a hard gate — the system cannot approve any section itself. Approved sections are locked and marked with a visual indicator. Every edit and approval is logged.
Finalize
When all sections are approved, finalize the workspace. The system runs a validation pass — checking cross-section consistency, unresolved placeholders, and critical contradictions — before locking the documentation set.
TD Workspace — Hyaluronan Soft Tissue Filling Gel
Workspace +
📂 Hyaluronan TD
📕 Your Work
📄 TD-1.1-01
📄 TD-1.2-01
📄 TD-2.1-01
📄 TD-3.0-01
📄 TD-4.1-01
📄 TD-5.1-01
📄 TD-6.1-01
Workspace Assets ↑
Process_Validation.pdf
Risk_Analysis_v3.pdf
Biocomp_Report.pdf
📄
TD-2.1-01 — Manufacturing Process Description
Draft
Device Description
2026-05-24
Draft — Manufacturing Process Description
Save Section
Edit Content
Back
Sebastian AI
How should I reference the sterilization validation in this section?
The manufacturing process description should reference the validated parameters from your process validation report. Focus on the critical process steps that directly affect device safety.
References and Warnings
High
📄 MDR Annex I, §11.1
📄 ISO 13485:2016 §7.5.6
⚠ Sterilization method not specified in evidence
Ask ReGentra... ➤
Template Coverage

54 templates across 8 documentation categories

Covering the full scope of MDR Annex II and Annex III requirements.

6 templates
Device Identification
Description, UDI, IFU, labelling, materials, SSCP
10 templates
Design & Manufacturing
Process, packaging, sterilization, facility, validation
1 template
GSPR
Checklist and evidence matrix
6 templates
Risk Management
Plan, report, controls, traceability, residual risk
9 templates
Verification & Validation
Biocompatibility, shelf life, usability, packaging, standards
8 templates
Clinical Evaluation
CER, literature, PMCF, equivalence, SSCP
4 templates
Post-Market Surveillance
PMS plan, PSUR, vigilance, trend reporting
10 templates
Administrative
DoC, PRRC, QMS certificates, UDI master list

8 categories covering the full scope of MDR Annex II and Annex III.

Regulatory Depth

What makes this a documentation engine, not a text generator

The mechanisms that separate structured regulatory production from general AI output.

Claim Classification & Traceability

Every generated sentence is classified as Normative, Derived, or Descriptive. Normative and derived claims are mapped to a persistent claim-evidence registry — a claim without a traceable evidence link is a system error. Normative claims without a source reference are blocked from generation.

187 Cross-Section Dependency Rules

54 templates are connected by 187 rules: 44 must-match (identical data required), 71 cross-reference (one section must cite another), and 72 logical alignment rules. A sterilization method stated differently across sections is caught automatically.

526-Entry Regulatory Knowledge Graph

The system draws from a structured reference database spanning EU MDR 2017/745 articles, ISO standards (13485, 14971, 10993 family, 62304, 60601), IEC, ASTM, and MDCG guidance documents. This graph drives cross-referencing during section generation, standard version validation, and evidence-to-requirement mapping.

Human-in-the-Loop by Design

Draft → Edit → Approve → Finalize. Human approval is a hard gate — the system cannot approve any section itself. Every generation event, edit, and approval is logged. This is an architectural decision, not a temporary constraint.

Workspace-Scoped Assistant

The assistant operates strictly within the active workspace — it only draws from that product's uploaded documents, never from other workspaces or general knowledge. Answers include source document, section, and page references, with confidence scoring (High/Medium/Low) per response.

Finalization Validation Gate

Before finalization, the system validates the entire document set. Sterilization contradictions, SAL mismatches, and unresolved critical placeholders block completion. Non-critical gaps produce advisory warnings without blocking.

Additional mechanisms include stale product data detection via hash comparison, cross-reference truncation alerts, zero-shot generation warnings for sections without uploaded evidence, and atomic state writes with full audit trail persistence.

54
TD Templates
187
Dependency Rules
12
Evidence Types
526
Regulatory References
100+
Standards Mapped

"Structured documentation, not generated text."

Evidence-grounded. Deterministic where it matters. Human-approved at every gate.

Get Notified

TD Workspace is currently in limited access. Leave your details to be notified when it becomes available.

Get Notified