Build a knowledge base
worth trusting.

Start from a library of textbooks and reference materials. Add your own documents, papers, and online sources. Aleph weaves them into a structured knowledge base — scored for reliability, honest about disagreements, with every claim traced to a source. Then connect it to Claude, Cursor, or any AI tool you already use.

AI tools are only as good as
what you give them to read.

Paste a few PDFs into a chat and the model treats everything equally — a peer-reviewed textbook and a blog post get the same weight. When sources contradict each other, it picks one at random or blends them into something neither actually said. And it never tells you where its knowledge is thin.

Aleph takes a different approach. Start from established reference materials — textbooks, primary sources, the kind of things you'd find on a university reading list. Layer in your own research and documents. The result is a knowledge base where every claim is scored, disagreements are mapped, and answers trace back to sources you actually trust.

Library to knowledge base in three steps

01

Browse the library

Start with our curated collection of textbooks, reference works, and primary sources — covering philosophy, science, game theory, literature, and more. Pick the foundations that matter for your field.

02

Add your own materials

Upload your own PDFs, research papers, and notes. Paste links to online references. Your materials get woven together with the library texts into a single, structured knowledge base — scored for reliability, with disagreements mapped honestly.

03

Use it everywhere

Connect your knowledge base to Claude, Cursor, or any AI tool that supports MCP. Ask questions and get answers grounded in real sources — textbook foundations plus your own research, with every claim cited.

Not just another document chatbot

Most tools dump your documents into a search index and hope for the best. Aleph starts from real reference materials and builds structured knowledge — the kind that holds up when you push on it.

Textbook depth, not chatbot guesses

Start from established references — the kind of material that gives AI real grounding. Then layer in your own research. The result is a knowledge base with actual foundations, not a pile of scraped web pages.

foundation: 3 textbooks · your docs: 12 · total topics: 87

Disagreements, not false certainty

When sources disagree — and serious sources often do — we don't pick a winner. Both sides are mapped with their evidence, so you get the real picture instead of a coin flip.

tension: active · sides: 2 · resolution: pending

Every claim traced to a source

Nothing appears from thin air. Every piece of knowledge links back to the textbook chapter, paper, or document it came from. You can always check the original.

source: Kahneman — Thinking, Fast and Slow ch.14 · type: reference

It knows what it doesn't know

Your knowledge base tells you where coverage is thin or based on weak evidence — so you know exactly which library text to add or what to upload next.

gap: long-term adherence · suggested: add clinical trial data

Use it inside the tools you already have

Aleph connects to Claude Desktop, Cursor, Windsurf, Zed, and any tool that supports MCP — the open standard for giving AI access to live data. No new app to learn. No extra tab to keep open.

Claude Desktop

Ask Claude questions grounded in your documents. It cites every source.

Cursor

Give your coding assistant access to internal docs, API specs, and standards.

Any MCP client

One open standard. Windsurf, Zed, and more — your wiki works the same everywhere.

Simple pricing

Library access is included. Pay once to build your knowledge base from library texts plus your own materials, or subscribe for a hosted connection you can query from any tool.

Wiki Build

From $5 one-time

Upload your docs, get a cost estimate, pay once. You own the wiki — download it anytime as plain files.

  • Reliability scores on every topic
  • Disagreements between sources mapped honestly
  • Every claim traced to the original document
  • Export as plain files anytime — you own it
Get an Estimate
Recommended

Hosted MCP

$20 / month

We host your knowledge base and keep it connected. Ask questions from inside Claude, Cursor, or any MCP-compatible tool. Includes ongoing sync when you add new docs.

  • Everything in Wiki Sync
  • Query your knowledge base from inside Claude, Cursor, or any MCP tool
  • Always-on connection — ask questions anytime, get sourced answers
  • We handle hosting, uptime, and updates
Get Started

Wiki Sync also available standalone at $15/mo. Custom builds available.

You own everything. Export anytime.

Your knowledge base is a folder of readable files — Markdown pages and a knowledge graph. Library texts, your uploads, all of it. Every plan includes export. Download it and use it however you want — drop it into Claude Projects, a custom GPT, or your own system. Nothing is locked behind our platform.

your-wiki/
├── wiki/
│   ├── graph.json
│   ├── concepts/
│   │   └── transformer_architecture.md
│   ├── positions/
│   │   └── intermittent_fasting.md
│   └── entities/
│       └── geoffrey_hinton.md
├── sources.json
└── metadata.json

Questions

What's in the library?

A curated collection of textbooks, reference works, and primary sources across a range of subjects — philosophy, cognitive science, game theory, economics, literature, mathematics, and more. Think of it as a university reading list you can draw from. We're expanding it continuously, and you can request additions.

How is this different from just pasting my docs into Claude?

When you paste documents into a chat, the model treats everything equally — a blog post gets the same weight as a peer-reviewed study. It can't tell you when sources disagree or how confident it should be. Aleph builds a structured knowledge base first: combining library texts with your own materials, scoring reliability, mapping conflicts, and tracing every claim to a source. The model gets organized knowledge instead of a wall of text.

Do I have to use the library texts, or can I just use my own stuff?

Entirely up to you. The library is there as a foundation — it gives your knowledge base depth and grounding that's hard to build from scratch. But you can skip it entirely and just upload your own documents, or start with library texts and add your materials later. Mix and match however you want.

What if my sources disagree with each other?

That's where Aleph is most different. Instead of quietly picking one side, we map both positions with their supporting evidence and flag the conflict. When you ask a question that touches a disagreement, you see both sides and the strength of each — not a false consensus.

Does this work with ChatGPT, Cursor, or other tools?

Aleph connects via MCP — an open standard supported by Claude Desktop, Cursor, Windsurf, Zed, and a growing list of AI tools. If your tool supports MCP, it works with Aleph. You can also export your knowledge base as plain files and use it anywhere, including Claude Projects or a custom GPT.

What do I actually get?

A structured knowledge base: Markdown pages organized by topic (concepts, entities, positions), a knowledge graph mapping how ideas connect, reliability scores on every topic, and a source manifest tracking where everything came from — both library texts and your own uploads. You can browse it, export it, or query it live from your AI tools.

Can I cancel or export anytime?

Yes to both. Every plan includes full export — your knowledge base as a zip of Markdown files plus graph.json. Cancel the subscription and you keep everything you've built. No lock-in, no exit fees.

A library of reference materials.
Your research on top.
One knowledge base that actually works.

Browse the library. Add your own sources. Get a knowledge base that scores reliability, maps disagreements, and connects to the AI tools you already use.

Explore the Library
For developers— MCP tools, config, and integration details

Connect in one line

Add this to your Claude Desktop or Cursor MCP config. That's it.

MCP config
{
  "mcpServers": {
    "my-knowledge-wiki": {
      "url": "https://aleph.lobstercollege.com/api/mcp/your-wiki-slug"
    }
  }
}

Five MCP tools

Your wiki exposes these tools to any connected MCP client. The model calls them like any other tool — no prompt engineering required.

query(question: string)

Search the wiki. Returns relevant pages with confidence scores, source attribution, and tension flags.

get_tensions(topic: string)

Find disagreements. Returns structured conflicts between sources on a given topic.

get_page(path: string)

Retrieve a full wiki page — metadata, markdown body, confidence, sources, and links.

list_sources()

View the source manifest with trust scores — study design, recency, rigor.

suggest_gaps(topic?: string)

Find where the wiki is thin or uncertain. Know exactly what to upload next.

Compatible with Claude Desktop, Cursor, Windsurf, Zed, and any MCP client.

21e9e8e