Genealogy Research Analyst - Claude Plugin Documentation
An AI-powered research partner for family historians using Claude.
Built by Denyse Allen / Chronicle Makers
What it does
The Genealogy Research Analyst turns Claude into a research partner that follows the Genealogical Proof Standard (GPS) — the same methodology used by professional genealogists and required by lineage societies like the DAR and SAR.
It adds five commands and three knowledge skills to Claude, so you can plan research, analyze sources, build proper citations, evaluate conflicting evidence, and write proof arguments — all with an AI that actually understands the methodology behind credible genealogy.
Commands
Command | What it does |
/research-plan | Creates a GPS-compliant research plan with prioritized record searches, repositories, and tracking |
/analyze-source | Classifies any genealogical source (original/derivative, primary/secondary, direct/indirect) and extracts every genealogically relevant detail |
/cite | Coaches you through building proper citations using the Jones five-question framework. It teaches you the skill — it doesn't just generate citations for you. |
/evaluate-evidence | Correlates multiple sources, identifies conflicts, and assesses which evidence is strongest. Honest about what's proven vs. what's claimed. |
/research-report | Generates GPS-compliant proof summaries, proof arguments, or full research reports. Refuses to write conclusions the evidence doesn't support. |
Skills (activate automatically)
Skill | When it activates |
GPS Methodology | When you ask about proof standards, exhaustive searches, or whether your research is "done" |
Record Types | When you ask what records exist for a time period, place, or research question |
Citation Guide | When you ask how to cite a source or document your research |
What makes this different
It follows the Genealogical Proof Standard
Every command is built around the five GPS elements: reasonably exhaustive search, complete citations, evidence analysis, conflict resolution, and soundly reasoned conclusions. This isn't a generic AI that happens to know some genealogy vocabulary. It applies the actual methodology.
It won't tell you what you want to hear
If your evidence doesn't support a conclusion, it says so. If your search isn't exhaustive enough, it tells you what's missing. If a source is derivative with secondary information, it explains why that matters for how much weight it carries.
It teaches, not just generates
The /cite command coaches you through building citations yourself using the Jones five-question framework. The /analyze-source command explains why a source is classified the way it is. You learn the methodology as you use it.
It knows what records exist — everywhere in the U.S.
The record-types skill includes era-specific guidance — Colonial, Revolutionary, Early Republic, Civil War, Gilded Age, and Modern — so it can tell you what to look for based on when and where your ancestor lived, in any state. It understands that record availability varies by era and jurisdiction: when vital records began, what each census year asked, what military records exist for each conflict, and which repositories hold what.
How to install
Claude Code
/plugin marketplace add DenyseCM/genealogy-research-analyst
/plugin install genealogy-research-analyst@chronicle-makers-genealogyClaude Desktop (Cowork)
Install via the Marketplace to the Github repository https://github.com/DenyseCM/genealogy-research-analyst through Cowork's plugin manager.
See it in action
Here's what the plugin does with a real research problem — a Revolutionary War militia captain whose death date is recorded differently in two lineage society applications.
The problem: The SAR application (1962) says Stephen Crumrine died in 1792. The DAR application (1998) says April 10, 1812. Federal census records show him alive in 1800 and 1810.
What /evaluate-evidence produced:
- Classified all five sources (SAR, DAR, 1800 census, 1810 census, tax records) by source type, information type, and evidence type
- Built a correlation table showing what each source says about each key fact
- Identified that the SAR application contains a provable error (mother's death date listed as 1725 — 12 years before the subject's birth) that passed review uncorrected
- Concluded the 1792 date is disproven by two original census records
- Assessed the 1812 date as plausible but unverified — honest about the gap
- Listed specific records still needed (probate, church burial, cemetery, deed records) before a GPS-compliant conclusion can be written
What /research-report did next:
- Refused to write a proof summary (evidence insufficient)
- Chose proof argument format instead (appropriate for unresolved conflicts)
- Wrote a 9-section document with full citations, evidence analysis, conflict resolution, and honest GPS assessment
- Left citation placeholders where microfilm numbers were missing rather than fabricating them
- Ended with prioritized next steps the researcher can act on
Want the full version? Join Chronicle Makers.
This free plugin gives you a strong foundation. But the extended version inside the Chronicle Makers community on Skool goes much further.
As a Chronicle Makers member, you get:
- Additional skills and commands not available in the public plugin
- Hands-on workshops where you learn to use every command with your own research
- Community support from fellow family historians who are actually finishing their stories
- Direct access to Denyse Allen for questions about methodology, tricky evidence problems, and getting unstuck
Chronicle Makers is where family historians stop researching in circles and start writing. The plugin gets you started. The community gets you finished.
About the author
Denyse Allen is a Pennsylvania genealogy researcher, AI builder, published author (Colonial Pennsylvania Genealogy Research, Pennsylvania Vital Records Research, Archives in Pennsylvania for Genealogy Research), and founder of Chronicle Makers — helping family historians transform research into published stories.
License
FSL-1.1-MIT — Free to use for research, education, and personal projects. Cannot be repackaged as a competing commercial product. Converts to MIT after two years.