Memoforge

memoforge
From one legal question to a finished .docx memorandum — researched, quoted, footnoted and stress-tested — in a single command.

~60-second walkthrough, recorded on v1.1.1: intake → research plan → parallel researchers → reviewer stress test → exported .docx. v2 is the same shape with fewer, better-checked steps.

What it does

memoforge turns a research-grade legal question into a structured, footnoted memorandum. It runs the pipeline a small legal team would run: an analyst clarifies the missing facts, researchers pull primary sources in parallel (statutes, case law, regulator guidance), a currency check confirms the law is still good, a writer produces an IRAC draft, reviewers stress-test it against fixed checklists, a mediator consolidates the findings, and the writer revises before export. It assumes no jurisdiction — the plan step classifies the query and routes research accordingly. EU data protection (GDPR, AI Act, NIS2, DSA), US privacy and sectoral regulation, UK consumer law and cross-border compliance are its strong ground.

What changed in v2 is where the pipeline lives. Phases, budgets, source freezing, citation provenance and every state.json write belong to a Python package (scripts/memoforge/, the mf CLI) with ~3,600 offline tests behind it. The orchestrating model follows one three-command protocol — next → act → report — and holds no pipeline knowledge that context summarisation can eat.

What you get

- memo-.docx — Arial 12pt, 1″ margins, numbered sections, IRAC per issue, in English, German, French, Spanish or Russian. Citations are OSCOLA-style: by default a short inline citation linked to the source, with the full record in a Sources annex; with citation_style: footnotes, *real* Word footnotes. Without python-docx the run still…