An address is a starting point. Evidence tells the story.

Plot & Kin helps professional researchers turn property-history questions into reviewable evidence dossiers. Start with an address and a question. Gather sources, trace people and places, compare conflicting accounts, and carry every conclusion back to its evidence.

An open-source MCP server and Codex workflow plugin, built for Codex and Claude Desktop. The service runs on your machine, research records live in your own Astra database, and original documents stay in your local library.

From a lead to a record you can inspect

• Evidence at the point of the claim. Link supporting and opposing passages, pages, and image regions. Less precise citations remain visibly less precise.

• A history that keeps its uncertainty. Separate proposals from approved conclusions, preserve competing explanations, and retain ambiguous or undated events in the timeline.

• Review that leaves a trail. Record decisions against the reviewed version. Corrections preserve prior extractions, and ambiguous identity merges require explicit review.

• Documents beyond searchable text. Extract embedded text or choose OpenAI or Anthropic processing for scans, handwriting, photographs, and map excerpts.

• Research with boundaries. Checkpoint progress, retain failed searches and next steps, and enforce run limits plus a default $10 project processing budget.

• Work you can take with you. Export Markdown, HTML, and JSON. Back up originals, citations, corrections, decisions, and logs, then restore into a fresh project.

flowchart LR
  A[Address + question] --> B[Sources + passages]
  B --> C[Proposed findings]
  C --> D[Researcher review]
  D --> E[Dossier + portable backup]
  D -->|Questions remain| B

Related MCP server: traceable-research-mcp

Try a complete example

Requires Node.js 22.19+. The synthetic demo needs no database credentials or model API key.

git clone https://github.com/jestatsio/plot-and-kin.git
cd plot-and-kin
npm ci
npm run build
node dist/cli.js demo

Open the HTML export at the path printed by the command. The demo also creates a portable backup and verifies a restore. Demo records are temporary, exported files persist, and no human approval is invented.

Ready to use your own sources? Follow the setup guide to configure Astra, connect either client, and optionally enable model processing.

First stop: Washington, DC

The first locality combines HistoryQuest building records, 1880 Sanborn map excerpts, and researcher-supplied documents, images, and public document URLs. Subscription material enters through permitted manual imports.

Compiled building information is identified as a research lead. Map excerpts retain attribution and extent. Address matches remain candidates for examination. These sources provide a starting point, with coverage and access limits documented in source coverage.

Your tools. Your research record.

No model key is required for embedded-text workflows. Model-assisted processing uses your chosen provider and credentials, with no silent provider switching. The $10 default processing cap excludes client charges and database hosting. See setup and recovery and evidence and data handling.

Built for a careful first pilot

Plot & Kin is a researcher-facing prototype under the Apache-2.0 license. Automated tests, real stdio checks, and live Astra/DC connector checks are recorded in validation. Actual desktop-client acceptance, live model extraction quality, and researcher effectiveness still need evaluation.

Review approvals are recorded workflow decisions, not an independently authenticated human-only control. Nationwide coverage, GIS alignment, shared accounts, hosted billing, and a standalone web application are outside this prototype.

Contributions to source coverage, evidence handling, and reproducible case evaluation are welcome. Start with the development checks.

Apache License 2.0 · Original materials retain their own rights and attribution requirements.