Skip to content
Product · How it works

An agent built for regulated writing.

Asthra is a four-step agentic loop, designed end-to-end for the constraints of regulatory submissions: closed-system retrieval, human-in-the-loop review, and audit-ready provenance on every claim.

Step 01 · Plan

Asthra writes a retrieval plan first — you approve it before it drafts a word.

Before generating any text, Asthra produces a plan: which source documents it intends to consult for each section of the deliverable. Writers review the plan, swap sources, exclude material out of scope, and approve.

  • Section-by-section retrieval mapping with source-file pinning

  • Edit, swap, or exclude sources before drafting begins

  • Plan is captured in the audit ledger and reproducible

Asthra · Retrieval plan
§9.1Subjects randomizedPROTOCOLTLF-DEM-01
§9.2Demographics & baselineTLF-DEM-01TLF-DEM-02
§9.3Primary endpointPROTOCOLSAPTLF-EFF-01
§9.4Adverse eventsTLF-AE-01TLF-AE-02
Asthra · Retrieving
PDFCSR-protocol-v3.pdf8 matches · 14 passagesIndexed
XLSTLF-DEM-01.xlsx3 tables · 24 cellsIndexed
DOCSAP-amendment-2.docx5 matches · 9 passagesIndexed
PDFprior-CSR-2023.pdf2 matches · 4 passagesIndexed
Step 02 · Retrieve

Closed-system semantic retrieval — your sources, only.

Asthra indexes the files you upload and pulls the right excerpts via semantic retrieval. No unsupervised internet access, no fallback to pretrained knowledge, no cross-tenant leakage — only the documents you specified. If an external lookup is genuinely needed, Asthra pauses and asks the writer to approve a specific request before fetching anything; the approval and the result are recorded in the audit ledger.

  • Section-aware semantic search, scoped to your document set

  • Tables, figures, and structured data ingested alongside narrative

  • Every retrieval returns a passage, page, and offset — recorded in the ledger

Step 03 · Draft

Section-specific drafting — never invented, always cited.

Claude drafts each section against module-specific writing instructions your team validated during onboarding. Every claim is bound to a retrieved passage. Where data is missing, Asthra surfaces an explicit gap flag rather than inventing plausible-sounding text.

  • Module-specific writing rules (ICH E3, EU MDR, ICH M4Q) applied per document type

  • Two-level citations: document-level by default, sentence-level on demand

  • Inline gap flags for missing or contradictory source data

CSR.docx · §9.4 Adverse Events
9.4   Adverse Events

Treatment-emergent adverse events were reported in 284 patients (88.5%) in the pembrolizumab arm and 295 (91.9%) in the docetaxel arm1.

Grade 3 or higher events occurred in 13.4% of pembrolizumab patients versus 35.0% of docetaxel patients2.

Subgroup analysis for patients aged ≥75 years ⚠ data gap — TLF-AE-03 missing

The most common immune-related events were hypothyroidism (8.1%) and pneumonitis (4.7%)4.

Audit ledger · KEYNOTE-189-CSR
14:32:01plan.create38 sections, 12 sources
14:32:14plan.approvewriter · m.singh@bayer.com
14:32:27retrievePROTOCOL · §5.1 → §9.1
14:33:02draft.write§9.1 · 247 tokens · 4 cites
14:33:48retrieveTLF-DEM-01 · row 2 → §9.2
14:34:11draft.write§9.2 · 312 tokens · 6 cites
14:34:55flag.gap§9.4 ≥75y · TLF-AE-03 absent
14:35:20handoffdocument → m.singh@bayer.com
Step 04 · Hand off

A draft you can audit — line by line.

The draft lands in Word with an immutable transaction ledger. Every plan change, retrieval, draft step, gap flag, and edit is recorded with a timestamp, actor, and target section. The ledger stays with the document — independent of Asthra — so audit readiness survives the vendor.

  • Append-only ledger embedded in the .docx as a side-channel attachment

  • Each citation resolves to a file, page, and exact passage

  • Track changes capture every writer edit alongside the agent record

Architecture

Built on Claude — wrapped in regulated-writing guardrails.

Foundation models alone don't ship in regulated environments. Asthra adds the orchestration layer, the closed-system retrieval, the citation graph, and the ledger that turn a capable model into an audit-ready writing agent.

Microsoft Word add-in
Native task pane. Writers stay in their existing authoring environment.
Asthra agent & orchestration
Plan → retrieve → draft → review loop. Module-specific writing rules. Citation graph. Audit ledger.
Closed-system retrieval (your VPC)
Semantic indexing of customer-uploaded documents. No cross-tenant data. Internet access is closed by default and approval-gated when the writer needs it.
Anthropic Claude
Frontier reasoning under structured prompts. Latest available model, via Anthropic enterprise API.

See Asthra
on your documents.

30-day pilot. We benchmark Asthra's output against your existing process — and show you the writer-hours saved, line for line.