Distillary

Turn any knowledge source into a navigable, shareable brain.

Demo Brain | Docs | Discord


Quick start

Try an existing brain (30 seconds)

Browse a published brain right now:

  1. Go to the Demo Brain
  2. Click any source’s root thesis to read the argument
  3. Follow [[wikilinks]] to drill deeper
  4. Check entity pages — their backlinks answer “what does this brain know about X?”

Or query it from your agent — fetch /agent.json and follow the navigation instructions.

Create your own brain (15 minutes)

curl -fsSL https://raw.githubusercontent.com/distillary/distillary/main/install.sh | bash
cd ~/distillary && claude

Then say:

Add ~/Downloads/my-book.epub to my brain

That’s it. Three commands. Claude runs 14 agents in parallel and builds your brain in ~15 minutes. Open brain/ in Obsidian to explore.

Add a second source

Add books/another-book.epub to my brain

Claude auto-bridges the new source to everything already in your brain — finds same concepts under different names, writes a comparison essay, creates unified entity pages.

Publish

Publish my brain

Your brain becomes a static website with graph view, search, backlinks, and an agent API at /agent.json.


What Distillary does

You give it a source (book, video, article). It gives you:

  • A pyramid of claims — root thesis → chapter clusters → argument groups → atomic evidence, each traceable to a chapter via source_ref
  • Entity hubs — every person, concept, and company gets a page where backlinks from every source show what your brain knows about it
  • Bridge concepts — when two sources call the same idea by different names, a bridge entity unifies them with both perspectives
  • Your annotations — your reactions, questions, and insights are part of the graph

All in one brain/ vault that works in Obsidian and publishes as a website.


How it works

Source (book, video, article)
  → 16 parallel haiku agents extract atomic claims (~2 min)
  → Deduplicate + extract entities (~2 min)
  → Opus agents build argumentative pyramid (~5 min)
  → Find tensions, patterns, evidence (~2 min)
  → Doctor fixes issues + suggests explorations (~1 min)
  → Bridge to existing sources in brain

14 agents total. Haiku for fast parallel work. Opus for reasoning. Everything defined in .claude/agents/ and orchestrated by .claude/skills/.


For AI agents

Published brains have an agent.json manifest. Any agent can query them:

curl https://brain.distillary.xyz/static/skill.txt   # get the skill
curl https://brain.distillary.xyz/static/agent.json  # get the manifest

Paste the skill into any agent’s system prompt — Claude Code, Codex, Gemini CLI, Cursor, or anything with HTTP access. The agent fetches agent.json, picks a strategy, and walks pages by relevance. 2 fetches, under 2000 tokens per question.

Demo: 10 questions answered from the live brain

#QuestionStrategyFetches
1What books are in this brain?Read manifest1
2What are the main ideas?Fetch clusters2
3What does Barnum say about debt?Concept lookup2
4Does luck play a role in success?Concept lookup2
5How important is integrity?Concept lookup2
6Career advice?Concept lookup2
7Role of health in wealth?Concept lookup2
8How to advertise?Concept lookup2
9Who is P.T. Barnum?Entity lookup2
10Root thesis?Fetch root note2

Average: 1.9 fetches per question. All answers sourced from brain.distillary.xyz — no local files. Full answers →

Five question strategies: concept (“what is X?”), source (“summarize”), comparison (“do they agree?”), evidence (“prove it”), exploration (“what’s related?”). Setup guide →


Community

Tag repos distillary-brain on GitHub. Join the Discord. Share brains, request distillations, compare sources.


Project structure

.claude/           17 agents + 10 skills (the brain)
distillary/        Python utilities agents call
brain/             Your knowledge vault (the output)
docs/              Documentation

Full docs → | Architecture → | Agent demo →


License

MIT


Built with Claude agents. Published with Quartz. The brains are the value, not the tool.