←Back/Product/Claude
IntermediateProduct

Build a Household Knowledge Base with Claude and Obsidian

Create a persistent, inspectable knowledge base for your household using Claude Code to process information and Obsidian to view the resulting Markdown files. This system acts as a reliable memory for appliances, warranties, and projects.

Illustration for Build a Household Knowledge Base with Claude and Obsidian

I wanted to build a system where I was confident that AI was grounded in my truth... I built this system that would read and write markdown files so they would have the memory of what we were talking about.

Before you start

What you need

  • A folder on your computer.
  • Unstructured information about your household (e.g., photos of appliance serial numbers, notes from brainstorming sessions, digital documents like warranties).

What you’ll make

A structured set of Markdown files in your folder, organized by topic (e.g., appliances.md, projects.md), which can be read by an AI and inspected in Obsidian.

Tools used

The workflow

Follow the sequence once, then adapt the prompts, checks, and handoffs to your own setup.

5 steps

Step 01

Create a Project Folder and Gather Initial Data

Create a dedicated folder on your computer. Gather initial household information. Scott Farrar started by asking Claude for suggestions, which led to taking pictures of every appliance's serial number plate.

Example prompt
I want to build a knowledge base for my house. What information should I gather first? What about my appliances?

Start small with one or two key areas, like appliances or a single ongoing project, rather than trying to document your entire house at once.

Step 02

Use Claude Code to Structure Information

Open your project folder in an AI coding agent like Claude Code. Provide it with your raw data (e.g., photos, unsorted notes) and ask it to organize the information into Markdown files.

Example prompt
Here is a folder of photos of my appliances. Create a Markdown file called 'appliances.md' that lists each appliance, its serial number, and any other identifiable information from the photos.

You can also provide a brain-dump of tasks from a brainstorming session and ask the AI to categorize and prioritize them in a 'todo.md' file.

Step 03

Instruct the AI on How to Maintain the Files

Provide instructions on how to update files, what format to use, and how to handle new information. This helps maintain consistency over time.

Example prompt
When I give you updates, you should append them to the relevant project file. Keep entries in the to-do list brief. Here are the sections I want in my 'todo.md' file: [Task, Priority, Cost Estimate, Reference].

Scott Farrar mentions that an earlier version became too verbose, so he had to add instructions to keep it brief. Setting these rules early helps.

Step 04

Inspect and Verify with Obsidian

Open your project folder as a 'vault' in Obsidian. Use Obsidian's interface to read the files, check their contents for accuracy, and see the connections between different notes.

Obsidian's graph view can help you visualize how different pieces of information (projects, appliances, contacts) are linked, confirming the AI is structuring the data correctly.

Step 05

Test the System with a New Chat Session

To confirm the 'memory' is working, start a completely new chat session with Claude Code. Ask a question that requires it to read the files you've created.

The goal is for the AI to retrieve the correct, household-specific information from your files without you having to paste it into the chat.

What good looks like

  • A new Claude Code session can answer a question like 'What's the warranty on my fridge?' by reading the files, without you re-providing the context.
  • You can see the organized information, like an appliance list, as a Markdown file in Obsidian.

Build your next product with ChatPRD

Turn an idea into a PRD, user stories, and a plan.

Try ChatPRD free

Runbook notes

How to recover when the loop fails and where human judgment helps.

If it goes sideways

The AI agent creates overly verbose or poorly structured files, making them hard to read and use.
Instruct the agent to be brief and follow a specific structure. Scott Farrar mentioned using a linting script to enforce rules, or simply starting a new chat session with clearer instructions if the agent gets off track.
The AI hallucinates or misremembers details from previous interactions, writing incorrect information into the files.
Regularly inspect the Markdown files in Obsidian to verify the information. Ask the AI 'Why did you think that?' to trace the source of the error. Manually correct any mistakes in the files.

Human checkpoints

Let routine work keep moving.

  1. Review the initial file structure proposed by the AI to ensure it makes sense for your household.
  2. Periodically inspect the generated Markdown files for accuracy and verbosity.

Start shipping better products.

Join 250,000+ product managers who use ChatPRD to write better docs, align teams faster, and build products users love.

Free to start · 14-day Pro trial · SOC 2 Type II