LIFA.VC LIFA MedTech Tools

MedTech AI Skills & MCP Index

A curated index of publicly available AI agent skills and MCP servers relevant to MedTech startups — devices, diagnostics, and digital health / SaMD.

Reviewed June 2026 · maintained quarterly

  • Useful to most MedTech founders
  • Digital health / SaMD
  • Diagnostics / genomics only
  • Bench-research (largely not founder-facing)
On this page

How to read this

Skill
— a Markdown file (plus optional scripts) that teaches an AI agent how to do a task. Procedural know-how loaded into context on demand.
MCP server
— a running service that gives an agent a live connection to an external system (a database, API, or pipeline) via standardised tool calls.
  • Cross-tool by default. Most skills here follow the open Agent Skills standard, so they work across Claude Code, Codex, Cursor, Gemini CLI, and others — not just Claude.
  • An index, not a code dump. Links go to upstream with the licence noted; nothing is re-hosted. Install from source.
  • Access. Skills need a paid Claude plan (Pro/Max/Team/Enterprise) inside Claude; MCP servers need a compatible client.
  • A blunt caveat. The life-science skill ecosystem is large but skews heavily toward academic bench research (RNA-seq, variant calling, ChIP-seq, population genetics) — mostly for computational biologists, not founders building a company. The genuinely founder-relevant slice is flagged below; start there.

Start here — by company type

The narrow, cross-cutting slice that actually helps a company get built and funded, by company type.

Every MedTech company

  • Clinical trial protocol drafting — clinical-trial-protocol (Anthropic life-sciences & healthcare; OpenClaw-Medical)
  • Regulatory submission drafting (FDA / EMA / ICH) — OpenClaw-Medical
  • Clinical / medical report writing — CSR per ICH-E3, CARE case reports — OpenClaw-Medical
  • Literature search & evidence synthesis — DeepMind science-skills; K-Dense; OpenAI life-science-research
  • Scientific writing & biostatistics — SciAgent-Skills; K-Dense
  • Device & drug regulatory intelligence — OpenFDA MCP (510(k), device classifications, adverse events, recalls)
  • Clinical trial landscape — ClinicalTrials.gov MCP / BioMCP
  • Quality system & SaMD compliance — ISO 13485 Certification (K-Dense) is the one substantial QMS skill; 21 CFR 820 / ISO 14971 / IEC 62304 tools remain early-stage (see Regulatory, QMS & SaMD compliance below)

Digital health / SaMD adds

  • FHIR development — Anthropic fhir-developer ; TopologyHealth ClaudeFHIRSkill
  • Prior authorisation & coverage (US market access) — Anthropic prior-auth-review , CMS coverage
  • FHIR + LOINC MCP — for interoperability

Diagnostics adds

  • Variant interpretation / federated variant query — ClawBio; DeepMind science-skills; BioMCP
  • Precision-oncology & rare-disease tooling — ToolUniverse
  • MyVariant / gene–disease–drug MCPs

Skills catalog

Official first-party skills

First-party repos — the highest-credibility sources.

Repo anthropics/skills
Relevant contents Document skills (docx, pdf, pptx, xlsx) + example skills (skill-creator, brand-guidelines, internal-comms, frontend-design, mcp-builder, webapp-testing)
Licence Example skills Apache-2.0; document skills source-available, not open source (cannot be re-hosted; ship inside Claude)
Repo anthropics/life-sciences
Relevant contents First-party skills: clinical-trial-protocol-skill, scientific-problem-selection, single-cell-rna-qc, scvi-tools, nextflow-development; bundled with partner MCPs
Licence Verify (marketplace repo)
Repo anthropics/healthcare
Relevant contents Skills: prior-auth-review, fhir-developer, clinical-trial-protocol, cms-coverage, icd10-codes, npi-registry, pubmed (consolidated into one healthcare plugin)
Licence Verify
Repo google-deepmind/science-skills
Relevant contents ~36 skills: genomics, structural biology, cheminformatics, literature search; wraps AlphaGenome, AlphaFold DB + 30 databases
Licence Apache-2.0
Repo openai/plugins — life-science-research
Relevant contents 50 modular skills: human genetics, functional genomics, pathway biology, protein structure, chemistry, clinical evidence
Licence Verify
Repo openai/skills
Relevant contents OpenAI's official Codex skills catalogue (system / curated / experimental)
Licence Verify

Founder-relevant community skills

Most community life-science libraries are bench-research tooling. The founder-relevant content concentrates in two repos — K-Dense and OpenClaw-Medical — with ClawBio and ToolUniverse adding a diagnostics-only layer. Listed below are the specific named skills worth pulling, not the repos. Licence by source: K-Dense = verify · OpenClaw-Medical = mixed / inherited (link-only, never re-host) · ClawBio = verify · ToolUniverse = verify.

Skill ISO 13485 Certification
What it does ISO 13485:2016 QMS docs — gap analysis, templates for 31 required procedures (Quality Manual, Medical Device Files), QSR→QMSR transition, EU MDR harmonisation
Source K-Dense
Fit Device QMS
Skill Regulatory submission drafting
What it does FDA / EMA / ICH document preparation
Source OpenClaw-Medical
Fit Regulatory
Skill Clinical trial protocol
What it does device/drug protocol + FDA submission docs
Source OpenClaw-Medical
Fit All
Skill Clinical Reports
What it does CARE case reports, ICH-E3 / CSR trial reports, diagnostic reports; HIPAA/FDA/ICH-GCP
Source K-Dense
Fit All
Skill Clinical Decision Support
What it does biomarker-stratified cohort + treatment-recommendation reports, GRADE grading
Source K-Dense
Fit Dx / clinical
Skill Scientific Writing
What it does IMRAD, CONSORT / STROBE / PRISMA reporting guidelines
Source K-Dense
Fit All
Skill Literature Review
What it does PubMed / bioRxiv / Scholar, PRISMA systematic-review templates
Source K-Dense
Fit All
Skill Statistical Analysis
What it does hypothesis testing, power analysis, effect sizes
Source K-Dense
Fit Trial design
Skill scikit-survival
What it does time-to-event / survival modelling (Cox, RSF) for clinical endpoints
Source K-Dense
Fit Clinical endpoints
Skill PyHealth
What it does clinical ML on EHR; ICD-9/10 / NDC / RxNorm / ATC code translation; fairness & calibration
Source K-Dense
Fit SaMD
Skill NeuroKit2
What it does ECG / PPG / EEG / EDA / EMG biosignal processing, HRV
Source K-Dense
Fit Wearable / SaMD
Skill SHAP
What it does model explainability (Shapley values) — AI/ML SaMD transparency
Source K-Dense
Fit AI/ML SaMD
Skill pydicom
What it does DICOM read/write + PHI anonymisation
Source K-Dense
Fit Imaging
Skill FHIR API development
What it does SMART on FHIR apps & resource endpoints
Source OpenClaw-Medical
Fit SaMD
Skill Prior authorisation workflow
What it does coverage cross-referencing
Source OpenClaw-Medical
Fit Market access
Skill ClinicalTrials.gov query
What it does trial search via API v2
Source OpenClaw-Medical
Fit All
Skill histolab / PathML
What it does whole-slide-image processing & computational pathology
Source K-Dense
Fit Dx (pathology)
Skill Database Lookup
What it does ClinicalTrials.gov, FDA, ClinVar, Open Targets, ChEMBL, USPTO (78 DBs)
Source K-Dense
Fit All
Skill DepMap / PrimeKG
What it does cancer dependency data + Precision Medicine Knowledge Graph
Source K-Dense
Fit Dx / oncology
Skill Federated variant query
What it does 9 databases (gnomAD, ClinVar, Open Targets, GTEx)
Source ClawBio
Fit Dx (genomic)
Skill Pharmacogenomics report
What it does CPIC, 12 genes / 31 SNPs / 51 drugs + FDA labels
Source ClawBio
Fit Dx (PGx)
Skill Polygenic risk / variant reports
What it does PGS Catalog
Source ClawBio
Fit Dx (genomic)
Skill Precision-oncology / rare-disease tools
What it does subset of 68 skills
Source ToolUniverse
Fit Dx / target
Skill Market Research Reports
What it does TAM / SAM / SOM, Porter's Five Forces, PESTLE
Source K-Dense
Fit Commercial (adapt)
Skill Research Grants
What it does NSF / NIH / DOE / DARPA proposals (US — not EIC/Horizon)
Source K-Dense
Fit Funding (US)

Excluded — bench-research, nothing founder-facing: swaruplab/operon (180+ protocols), GPTomics/bioSkills, HKU-BAL/Clair-skills — RNA-seq, ChIP-seq, ATAC-seq, WGS/WES, spatial, GWAS, variant calling. Relevant only to a diagnostics company running NGS analysis in-house, and even then as R&D tooling, not company-building. jaechang-hits/SciAgent-Skills (197, CC-BY-4.0) is mostly the same bench-tool wrappers; its only founder-relevant skills (biostatistics, scientific writing) are covered better by K-Dense. Licences for K-Dense / ClawBio / ToolUniverse still need confirming before any reuse.

Regulatory, QMS & SaMD compliance

The one reasonably substantial QMS skill is K-Dense’s ISO 13485 Certification (listed above) — gap analysis, templates for the 31 required procedures, QSR→QMSR transition, and EU MDR harmonisation. Beyond that, this layer is thin: a cluster of early-stage, mostly single-author, near-zero-adoption tools under the GitHub `medical-device` topic. Worth tracking, but treat as unvetted: low adoption, "human review required," quality not established. Note: since the FDA’s QMSR incorporates ISO 13485:2016 by reference from 2 February 2026, a 21 CFR 820 tool effectively covers the US ISO 13485 angle.

Repo LSaiko/sop-review-tool
What it does Claude-based 21 CFR 820 SOP compliance reviewer — preliminary QMS gap analysis (human review required)
Licence Verify
Repo LSaiko/SaMD-Val-Kit
What it does FDA/IEC SaMD validation packages — IQ/OQ/PQ, ISO 14971 risk management, IEC 62304 lifecycle, SBOM, cybersecurity
Licence Verify
Repo LSaiko/pfmea-template
What it does AIAG-VDA 2019 pFMEA generator (CNC, injection moulding, sterile packaging) citing 21 CFR 820 / ISO 13485 / ISO 11607
Licence Verify
Repo CSOAI-ORG/fda-samd-mcp
What it does FDA AI/ML SaMD compliance MCP — 510(k), De Novo, PMA, PCCP, GMLP
Licence MIT
Repo CSOAI-ORG/mdr-medical-device-mcp
What it does EU MDR (2017/745) + IVDR + AI/ML SaMD classification MCP
Licence MIT
Repo masalu0105-gif/tfda-license-assistant
What it does Claude Code skill — Taiwan TFDA device licence / QSD/QMS open data
Licence Verify
Repo Tasdax123/claude-fidelity-translator
What it does Claude skill for safety-critical / regulated-document translation — terminology consistency, back-translation (useful for IFU & labelling across Nordic languages)
Licence Verify
Repo RAPath/rapath.github.io
What it does Plain-English medical-device regulatory reference (FDA, MDR, IVDR, 510(k), UDI, CE, MHRA) — a reference site, not a skill
Licence Verify

Reality check: this is the layer device founders most need and the layer that is least mature publicly. These are worth listing so founders know they exist, but none is a substitute for qualified regulatory-affairs work.

Clinical & commercial single-purpose

Repo / skill TopologyHealth/ClaudeFHIRSkill
What it does FHIR system-building across R4/R4B/R5 — FSH/SUSHI authoring, validation, multi-language server patterns
Licence "As-is" — verify
Fit
Repo / skill Biostatistics-in-R Claude Code marketplace
What it does 30 agents, 17 commands, 45 skills — Bayesian modelling (Stan/PyMC/JAGS), indirect treatment comparison (NMA/MAIC/STC), clinical trial simulation
Licence Verify
Fit 🟢 relevant for trial design / HTA
Repo / skill OneWave-AI/claude-skills — pitch-deck-reviewer
What it does VC-grade slide-by-slide deck review
Licence Verify
Fit 🟢 general startup — adapt for MedTech narrative
Repo / skill Fundraising Knowledge skill
What it does Term-sheet economics, stage metrics, investor outreach
Licence Verify
Fit 🟢 general startup

MedTech-fit warning: general-startup financial/pitch skills bake in SaaS assumptions (MRR/ARR, Rule of 40, the standard pitch arc) that mislead in MedTech’s milestone-driven, capital-intensive economics. Treat as scaffolding only.

Standards & methodology

Repo agentskills/agentskills
What it is The open Agent Skills specification — the cross-platform format everything above uses
Licence —
Repo obra/superpowers
What it is Software-development methodology skills (TDD, spec-driven design); improves reproducibility in any pipeline
Licence —
Repo github/awesome-copilot
What it is General skill collection
Licence —
Repo Prat011/awesome-llm-skills
What it is General skill collection
Licence —

MCP servers

The live data connections. Several founder-critical taps (device 510(k) data, clinical trials, literature) exist only as MCP, not as skills.

Regulatory & FDA

Server Augmented-Nature/OpenFDA-MCP-Server
Coverage openFDA: drug adverse events, labeling, recalls, approvals, NDC — and medical device 510(k) clearances, classifications, device adverse events
Licence Verify
Server CDISC Library MCP (via awesome-healthcare-mcp-servers)
Coverage CDISC clinical data standards for regulatory submissions
Licence Verify
Server Pharma R&D Pipeline Intelligence MCP
Coverage Clinical trials, FDA/EMA approvals, safety data, orphan drugs
Licence Verify

Clinical trials & evidence

Server aafjes/mcp-clinicaltrials.gov
Coverage ClinicalTrials.gov API v2 — 400k+ trials, 220+ countries
Licence "As-is" — verify
Server genomoncology/biomcp
Coverage Clinical trials, scientific literature, genomic variants in one server; NCI + gene/disease/drug tools
Licence MIT
Server Refua clinical-evidence-synthesis MCP
Coverage Scored access to ClinicalTrials.gov + PubMed + openFDA in one call (REST + MCP, no key)
Licence Verify

Literature & paper search

Dedicated paper-search servers. medRxiv & bioRxiv servers are listed across the curated indexes below, and several aggregators above already bundle PubMed.

Server MCP Simple PubMed
Coverage PubMed via the Entrez API — keyword / MeSH / author / date / Boolean search, abstracts & metadata, full text for open-access (PMC); systematic-review & PICO prompts
Licence MIT
Server BGPT
Coverage Scientific papers with extracted raw experimental data — methods, results, sample sizes, quality scores, 25+ metadata fields per paper; also a K-Dense bgpt-paper-search skill
Licence Freemium
Server Consensus
Coverage 200M+ peer-reviewed papers (Semantic Scholar, PubMed, arXiv) with study-design / SJR quartile / sample-size / human-only filters
Licence Freemium

Multi-API aggregators

Server pascalwhoop/medical-mcps
Coverage 100+ tools across 14 APIs — ClinicalTrials.gov, PubMed/PubTator3, OpenFDA, MyVariant, BioThings (MyGene/MyDisease/MyChem), NCI trials
Licence Verify (hosted endpoint available)
Server Cicatriiz/healthcare-mcp-public
Coverage FDA drug info, PubMed, medRxiv, NCBI Bookshelf, clinical trials, ICD-10, DICOM, medical calculator; one-click DXT install
Licence Verify
Server JamesANZ/medical-mcp
Coverage FDA, WHO, PubMed, Google Scholar, RxNorm; medical journal & guideline search
Licence MIT
Server biocontext-ai/knowledgebase-mcp
Coverage Open Targets, Reactome, UniProt, HPA, KEGG, AlphaFold, Ensembl, ClinicalTrials.gov, bioRxiv
Licence Verify (OSI-approved per registry criteria)

Genomics / variant (diagnostics)

Server longevity-genie/gget-mcp
Coverage Pachter Lab gget toolkit — Ensembl gene search, BLAST/BLAT, ARCHS4 expression, PDB/AlphaFold, COSMIC
Licence Verify
Server cafferychen777/ChatSpatial
Coverage Spatial transcriptomics (Scanpy/Squidpy) — ⚪ bench-research
Licence Verify
Server MyVariant / BioThings (via medical-mcps)
Coverage Variant annotations, population frequencies, clinical significance
Licence Verify

Official Anthropic connectors (adjacent — not open source, but worth knowing). PubMed · ClinicalTrials.gov · Medidata (trial feasibility / site performance) · BioRender · Synapse · Wiley Scholar Gateway · CMS Coverage Database · ICD-10 · NPI Registry · 10x Genomics · ChEMBL · Open Targets · ToolUniverse. Available through Claude for Healthcare / Life Sciences.

Curated indexes & awesome lists

Re-scan these to track the ecosystem.

List GoekeLab/awesome-genomic-skills
Scope Genomics/bioinformatics skills, MCPs, benchmarks across all agents
Licence CC0-1.0
List rdmgator12/awesome-healthcare-mcp-servers
Scope Healthcare MCP servers — FHIR, clinical decision support, revenue cycle, HIPAA; notes CMS-0057-F / HTI-1 mandates
Licence Verify
List AgenticHealthAI/Awesome-AI-Agents-for-Healthcare
Scope Healthcare agents incl. 51 US healthcare-administration agents with real 42 CFR / CMS citations
Licence Verify
List InternScience/Awesome-Scientific-Skills
Scope Bioinformatics, cheminformatics, scientific writing, literature search (plans a clone-ready repo)
Licence Verify
List biocontext-ai/registry
Scope Community-curated biomedical MCP catalogue; submission criteria require biomedical focus + OSI licences
Licence Verify
List MCPmed
Scope Reference biomedical MCP implementations + cookiecutter template
Licence Verify
List punkpeye/awesome-mcp-servers
Scope General MCP list with a Biology/Medicine/Bioinformatics section
Licence —
List travisvn/awesome-claude-skills
Scope General Claude skills index (official + community foundation)
Licence —

Licensing at a glance

These notes matter if you ever copy a skill into your own project:

Licence Apache-2.0
Seen in DeepMind science-skills; Anthropic example skills
What it requires when copied Keep LICENSE + NOTICE; state changes
Licence MIT
Seen in operon protocols; BioMCP; JamesANZ/medical-mcp; CSOAI SaMD & MDR MCPs
What it requires when copied Keep the copyright + licence line
Licence CC-BY-4.0
Seen in SciAgent-Skills
What it requires when copied Attribute per skill
Licence CC0-1.0
Seen in awesome-genomic-skills
What it requires when copied No conditions
Licence Source-available (not OSS)
Seen in Anthropic document skills
What it requires when copied Cannot re-host; use in place
Licence Mixed / inherited
Seen in OpenClaw-Medical (872-skill aggregator)
What it requires when copied Treat as link-only; per-upstream review before any reuse

"Verify" entries above have no licence confirmed yet — read the upstream LICENSE before any reuse.

Does curation help?

Independent benchmarks measuring whether curated skills improve agent performance:

  • SkillsBench — Across 7,308 trajectories, curated skills lifted average task pass rate by +16.2 points, and +51.9 points for healthcare specifically; self-generated skills gave no average benefit, and focused 2–3 module skills beat comprehensive documentation. repo →
  • BixBench — Real-world computational-biology agent tasks; the benchmark SciAgent-Skills tops at 92%. repo →
  • BioAgent Bench, CompBioBench (Genentech), LAB-Bench, BiomniBench, BioMed-AQA — Further bioinformatics agent evaluations.

Using these safely

Data integrity

These are drafting and analysis aids, not sources of truth. Anything that emits a statistic, market figure, or citation must be traced to a named primary source before it enters a deck, submission, or board paper.

Skills run code

A skill can execute arbitrary code in the agent’s environment. Review the SKILL.md and any scripts before installing — especially from community repos.

Regulatory & clinical limits

Regulatory submissions, QMS documents, and clinical protocols generated here need qualified regulatory-affairs and clinical review before use. These tools do not replace that, and are not medical advice.

MedTech fit

Most libraries here are academic bench-research; the founder-relevant slice is in “Start here” above. SaaS-shaped financial/pitch skills do not fit MedTech economics.