A compound knowledge system that grows itself.
Obsidian vault Β· Knowledge graph Β· Hermes Agent automation
A self-growing second brain built on four components:
| Component | Role | |
|---|---|---|
| π | Obsidian | Markdown knowledge base (vault) |
| π | Graphify | Wikilinks β interactive knowledge graph |
| π§βπ» | Claude Code | Director β orchestrates complex tasks |
| π€ | Hermes Agent | Executor β collects, ingests, automates via cron |
Key idea: Claude Code delegates, Hermes executes. The system runs 24/7 on a server β accessible from Discord or CLI.
Clone to ~/system/ and the directory is the system. No extra config needed.
hermes-second-brain/
βββ second-brain/ # Obsidian vault
β βββ CLAUDE.md # Claude Code instructions
β βββ .hermes.md # Hermes project context
β βββ raw/ # Raw materials (never modify)
β β βββ inbox/ # Unprocessed items
β β βββ articles/ # Web articles
β β βββ papers/ # Research papers
β β βββ notes/ # Manual notes
β βββ wiki/ # Curated knowledge base
β β βββ concepts/ # Concept documents
β β βββ entities/ # Entity documents
β β βββ sources/ # Source summaries
β β βββ synthesis/ # Integrated analysis
β βββ graphify-out/ # Knowledge graph output
βββ scripts/
β βββ wiki-graph.py # Wiki β graph generator
β βββ wiki-lint.py # Consistency checker
βββ hermes/skills/ # Hermes Agent skills
βββ kor/README.md # νκ΅μ΄ λ²μ
# 1. Clone
git clone https://github.com/Burgunthy/hermes-second-brain.git ~/system
# 2. Knowledge graph
python3 ~/system/scripts/wiki-graph.py
# 3. Hermes Agent
python3 -m venv ~/system/.venv && source ~/system/.venv/bin/activate
pip install git+https://github.com/NousResearch/hermes-agent.git playwright
npx playwright install chromium
# 4. Copy skills
cp -r ~/system/hermes/skills/* ~/.hermes/skills/Open ~/system/second-brain/ as an Obsidian vault and you're set.
π§ Full setup (systemd service, Claude Code, env vars)
second-brain/CLAUDE.md tells Claude Code to delegate to Hermes:
cd ~/system/second-brain
claudeEdit ~/.hermes/config.yaml:
model.defaultβ your preferred modelterminal.cwdβ~/system/second-brain
export DISCORD_BOT_TOKEN="***"
export OPENAI_API_KEY="***"# Create ~/.config/systemd/user/hermes-gateway.service, then:
systemctl --user daemon-reload
systemctl --user enable --now hermes-gateway Collect Ingest Query
ββββββββββββ ββββββββββββββββ ββββββββββββ
β Articles ββββ β Extract β β β
β Papers β ββββββΆ concepts ββββββββΆβ Wiki β
β Notes ββββ β entities β β Search β
ββββββββββββ ββββββββ¬ββββββββ ββββββββββββ
β
ββββββββΌββββββββ
β Knowledge β
β Graph β
ββββββββββββββββ
Ingest pipeline: raw/ β extract concepts & entities β wiki/ β update index & log β refresh graph.
Every wiki document uses this frontmatter:
---
title: Document Title
type: concept | entity | source_summary | synthesis
related: "[[RelatedDoc1]] [[RelatedDoc2]]"
sources: raw/file-path.md
created: 2026-04-22
updated: 2026-04-22
---Cron jobs run automatically via Hermes Agent:
| Schedule | Job | Description |
|---|---|---|
| Daily 4 AM | daily-ingest |
Scan raw/inbox/ β ingest to wiki β refresh graph |
| Sun 5 AM | weekly-lint |
Run wiki-lint.py consistency checks |
| Mon 9 AM | weekly-summary |
Summarize last week's wiki changes |
hermes chat -q "create cronjob: daily-ingest, 0 4 * * *, ingest raw/inbox/ to wiki/ and refresh graph, deliver: discord"raw/β never modify (preserve originals)wiki/β log all changes inwiki/log.mdwiki/index.mdβ always keep up to date
νκ΅μ΄ README Β· Built with Hermes Agent