obsidian-second-brain: a self-maintaining Obsidian vault that rewrites itself and synthesizes knowledge using LLM commands

obsidian-second-brain: a self-maintaining Obsidian vault that rewrites itself and synthesizes knowledge using LLM commands

What it solves

This project addresses the "institutional amnesia" that occurs when using AI assistants like Claude, where every session starts from scratch and knowledge is lost. It also solves the problem of static note-taking in Obsidian, where information often sits unused and disconnected. By bridging the two, it creates a "second brain" that doesn't just store information but actively maintains, updates, and synthesizes it.

How it works

It functions as a cross-platform skill (compatible with Claude Code, Codex, Gemini, OpenCode, and Hermes) that allows an LLM to interact directly with an Obsidian vault. Instead of simply appending new notes, the system rewrites existing pages to update facts and reconcile contradictions. It uses a layered architecture:

  • Operations: 28 commands for saving, ingesting content (URLs, PDFs, audio, screenshots), and organizing the vault.
  • Thinking Tools: 7 commands for challenging ideas using vault history, surfacing unnamed patterns, and distilling information.
  • Context Engine: A command to load the user's identity and state into the AI session.
  • Research Toolkit: 7 commands that integrate external data from X (via Grok), the web (via Perplexity), and YouTube/podcasts into the vault.
  • Background Agents: Scheduled agents that run nightly to reconcile contradictions, synthesize patterns, and heal orphan notes.

Who it’s for

Knowledge workers, researchers, and developers who use Obsidian for note-taking and want an AI-driven system that automatically manages their knowledge base, tracks decisions, and performs deep research without manual organization.

Highlights

  • Self-Maintaining Vault: Automatically rewrites pages and resolves contradictions rather than just appending data.
  • Multi-Source Ingestion: Converts voice memos, screenshots, and URLs into structured knowledge across multiple vault pages.
  • AI-First Formatting: Uses a specific preamble and frontmatter designed for LLM retrieval rather than human reading.
  • Codebase Architecture: Includes a /obsidian-architect command that scans codebases to maintain architecture notes in the vault.
  • External Integration: Live research capabilities via Grok and Perplexity to fill knowledge gaps in the vault.

Sources