Obsidian Research Assistant

The Problem

Research is overwhelming. You read papers, take notes, chase references - and six months later you’re staring at scattered files with no clear structure. The knowledge never compounds because it’s never connected.

Most AI tools make this worse: they answer questions but leave you with nothing persistent. You’re back to zero every session.

The Solution

The Obsidian Research Assistant transforms research from chaotic to systematic. You describe what you’re researching, and AI personas build your knowledge base - creating structured markdown files in your Obsidian vault with proper citations, linked concepts, and maps of content.

The key difference: you end up with artifacts, not answers. Every research session produces files that persist, connect, and grow.

How It Works

# Create a research vault (one-time setup)
./setup-vault.sh ~/vaults/my-dissertation "Transport Planning Research"

# Start researching
cd ~/vaults/my-dissertation
claude

# Activate Laura (research persona)
/laura

# Ask for research
"Research the fundamentals of transit-oriented development.
This is my dissertation so I need peer-reviewed sources."

Files appear immediately in Obsidian:

Everything linked. Everything yours.

The Personas

Three expert perspectives, each with distinct expertise:

Laura (Research Assistant) - Conducts systematic research, processes papers, builds concept notes with proper academic citations. Prioritises peer-reviewed sources for dissertations, adapts for professional contexts.

Alex (Solution Architect) - Evaluates technology options, creates Architecture Decision Records, assesses risks and trade-offs. Builds on Laura’s research to make technical decisions.

Riley (Product Owner) - Writes user stories, defines value propositions, prioritises features. Ensures research connects to user needs and business value.

Each persona reads what the others have built. Laura researches event streaming platforms; Alex evaluates the options and creates an ADR; Riley ensures the decision connects to user value. Same vault, multiple perspectives.

What Gets Created

Concept notes - One idea per file, with definition, context, examples, and links to related concepts.

Source notes - Full citations, methodology assessment, key findings, and links to concepts discussed.

Question notes - Research gaps and contradictions that need investigation.

Maps of Content - Thematic overviews that connect concepts into navigable structures.

Architecture Decision Records - Technology choices with context, alternatives considered, and consequences documented.

User stories - Requirements written from user perspective with testable acceptance criteria.

The Philosophy

AI augmentation, not automation.

Laura doesn’t write your dissertation. Alex doesn’t make your architecture decisions. Riley doesn’t prioritise for you. They handle the scaffolding - the structure, the connections, the organisation - so you can focus on thinking and understanding.

The knowledge base is yours. The insights are yours. The writing is yours.

Who It’s For

Graduate students building dissertation knowledge with academic rigour. Laura prioritises peer-reviewed journals, notes methodology, and creates proper citations.

Professional researchers getting up to speed fast in unfamiliar domains. When you’ve got a week to become the expert in the room, systematic knowledge-building beats scattered reading.

Technical teams making architecture decisions. Alex turns research into structured evaluations and documented decisions.

Product teams connecting research to user value. Riley ensures technical decisions trace back to user needs.

Tech Stack

No backend. No subscriptions. Everything runs locally and you own all the files.

Design Decisions

Skills over prompts - Personas are Claude Code skills (SKILL.md files) that auto-load when you invoke /laura. Rich methodology, not just personality.

Per-vault installation - Skills live in each vault’s .claude/skills/ directory. Portable and self-contained.

Artifacts over answers - Every interaction produces files. If you’re just getting chat responses, something’s wrong.

Progressive disclosure - Start with Laura for research. Add Alex when you need architecture decisions. Add Riley when you need product thinking. Each persona builds on the others.

What I Learned Building This

Skills > prompts - A comprehensive SKILL.md with methodology, templates, and quality checks beats clever prompting every time. Claude follows detailed instructions reliably.

The file creation mandate - Early versions would describe what files to create instead of creating them. Adding explicit “you MUST create files” directives fixed this completely.

Cross-persona collaboration - The real value emerges when personas read each other’s work. Laura’s research becomes Alex’s input; Alex’s decisions become Riley’s constraints.

Launcher scripts are optional - With skills working properly, /laura is enough. The friendly welcome banners are nice but not essential.

Status

Production ready for research and architecture work. Three personas (Laura, Alex, Riley) fully implemented. Used for real dissertation research and technical investigations.

Future personas planned: Morgan (Tech Lead), Quinn (Business Analyst), Sam (Scrum Master).

Built in February 2026 as an experiment in multi-persona knowledge building. Tested on a real transport planning dissertation and multiple technical investigations.

If you want more control, you can run your own instance from the source, or get in touch about a private setup.

View the source and install instructions on GitHub