mcpField
V
Vahtian

Vahtian

Local
github.com/heidihelenaScience & Education1Updated 12w ago
Repository

About Vahtian

Vahtian is a suite of research citation-integrity and systematic-review tools for researchers and research organizations. It records whether manuscript claims are supported by cited evidence through human ratings, AI advice, adjudication, and a hash-chained audit trail; it does not determine scientific truth.

Key features

  1. 01Check claims against their cited sources
  2. 02Use human, AI, and adjudication stages
  3. 03Gate and allow undoing of recorded decisions
  4. 04Maintain a hash-chained audit trail
  5. 05Provide systematic-review screening and extraction tools

Use cases

  1. 01Check whether paper claims have cited support
  2. 02Record reviewer judgments and dispute adjudication
  3. 03Run systematic-review screening and inter-rater analyses
  4. 04Extract diagnostic-accuracy data and assess bias risk

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
  "mcpServers": {
    "heidihelena-vahtian": {
      "command": "uvx",
      "args": [
        "vahtian"
      ]
    }
  }
}
Run commanduvx vahtian

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
pypivahtian0.2.0

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

vahtian

DOI

Company site for Vahtian — research citation integrity.

Vahtian checks whether each manuscript claim is supported by the evidence cited for it — through a blinded human → AI → adjudication workflow with a decision-gated, undoable write and a hash-chained audit trail.

Vahtian does not decide scientific truth. It records whether a cited source supports a specific claim: who rated it, what the AI advised, how disagreements were adjudicated, and what was finally written.

Human-first. AI-second. Auditable. Free local tools for researchers; infrastructure for journals, guideline groups, and research organizations.

The Vahtian tool family: Start (CiteVahti — check your own manuscript), Prepare (StudyVahti, MethodVahti, DictVahti, MatchVahti-Lite — get the study right first) and Scale (ReviewVahti, ExtractVahti, GuidelineVahti, AtlasVahti — systematic-review infrastructure), all standing on one evidence record: human rates first, blinded AI second, adjudicated, hash-chained, provenance or flag. Records whether a cited source supports a specific claim — never scientific truth.

The product ladder

Start — check the claims in your own manuscript

Prepare — get the study and its evidence right before analysis

  • StudyVahti — plan a defensible study before data collection (free · live): /studyvahti
  • MethodVahti — justify your sample size → COREQ/SRQR methods PDF (free beta for individuals, project licence for institutions): /methodvahti
  • DictVahti — clean the REDCap data dictionary before analysis (free · live): /dictvahti
  • MatchVahti-Lite — capture citation-worthy evidence from abstracts (free · live): /matchvahti-lite

Scale — systematic-review infrastructure

  • ReviewVahti — inter-rater reliability for screening (Cohen's κ / Krippendorff's α) (live): /reviewvahti
  • ExtractVahti — DTA data extraction + risk-of-bias (QUADAS-2/-C, PROBAST-AI) → tidy CSV + RoB traffic-light SVG for R; multi-extractor reconcile (κ / α); sealed agent aid (early access): /extractvahti
  • GuidelineVahti — panel consensus, RAND/UCLA + GRADE (early access): /guidelinevahti
  • AtlasVahti — de-identified validation-pattern analytics (roadmap)

Static pages (one index.html per product, plus 404.html and sitemap.xml), deployed via Cloudflare to vahtian.com. No trackers, no external dependencies.

vahti (Finnish) = sentinel / guard — the watch over what gets cited.

Brand & social tooling

  • brand/STYLE.md — the canonical brand spec; brand/PHILOSOPHY.md — the mark system.
  • brand/cards/ — 1200×630 Open Graph cards, one per page. Regenerate with npm i @resvg/resvg-js && node brand/cards/generate.mjs.
  • Screenshot-led product covers are generated from their real interface previews with node brand/cards/generate-product-covers.mjs (uses the bundled preview browser).
  • brand/carousel/ — self-contained LinkedIn carousel builder (noindex, zero external requests). Preset decks per product, theme & format switching, post-caption drafts, deck save/load as JSON, export as PDF (LinkedIn document post) or per-slide PNGs.

MethodVahti package

methodvahti/ is a self-contained Python package (Apache-2.0) for qualitative heterogeneity scoring + a COREQ/SRQR methods-report PDF. See methodvahti/README.md. Tests: cd methodvahti && python -m pytest tests/ -v.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.