Browse / Second brain / claude-obsidian-kit

claude-obsidian-kit

selmakcby

Give your Claude Code agents a shared, persistent brain — an Obsidian vault that survives sessions.

Open selmakcby/claude-obsidian-kit → Checked against the four rules.

ai-agentsanthropicclaude-codeknowledge-managementmarkdownobsidiansecond-brainvaultzettelkasten

Does not say

no agent config and no instruction file at the root.

In its own words

Aboutselmakcby/claude-obsidian-kit

Includes a retrofit guide for folders you already have, and a runnable vault linter.

README.md

Give your Claude Code agents a shared, persistent brain.

Turn ephemeral sessions into accumulating knowledge — stored in an Obsidian vault. Built for efficiency: right model for each job, cheap writes, index-first reads.

What this is

A drop-in kit that extends Claude Code with:

  • Two agents:
  • archivist — vault reader + health guardian (QUERY, LINT)
  • scribe — dedicated vault writer (INGEST, runs on haiku for efficiency)
  • One skill — llm-wiki — the INGEST · QUERY · LINT protocols
  • One constitution — vault-template/CLAUDE.md — the schema every agent reads before touching the vault (🇹🇷 Türkçe)
  • One vault template — pre-structured Obsidian vault with conventions
  • One linter — vault-lint.py, the LINT protocol as a runnable command (no dependencies)
  • One retrofit guide — RETROFIT.md, for when you already have months of material
  • Obsidian config — graph colours by layer, and the [!conflict] callout the skill needs

After install, your agents read from and write to a shared markdown knowledge graph that survives sessions. Yesterday's decisions are tomorrow's context.

The problem this solves

Default Claude Code agents are stateless:

Every session starts from zero. Context dies with the conversation. The project's reasoning evaporates even when the code persists.

The repository's own description, then 28% of its README. Read the rest on GitHub ↗

Read from the repository's git tree on 2026-08-17.