Skillbooks are structured, navigable knowledge graphs — served to AI agents on demand, per page, with citations. No hallucinations. No stale training data. No prompt bloat.
Free to browse · Pay only for pages fetched · 80% revenue to authors
Training data, web search, and flat skill files all fail agents in different ways. Skillbooks fix all three.
Training data freezes at a cutoff date and costs millions to update. A skillbook is a live repo — push a commit, every agent gets the update instantly.
Web search returns pages that mention a topic. Skillbooks return the exact page that answers the question, with regulatory citations, worked examples, and navigation to adjacent concepts.
Traditional skills dump entire documents into the prompt. Skillbooks let agents navigate a knowledge graph, fetching only the nodes they actually need — keeping context windows clean.
Every node is a URL. Every answer can point back to its source page. No more "I think I read somewhere..." — just clean, verifiable links agents can share with users.
A skillbook root is itself a skill — agents load it from their toolchain and start navigating. No integration work. No SDK. Just HTTP.
Microdollar per-page pricing. A full traversal of a complex technical book costs less than a cent. Authors earn 80–90% of every fetch.
Every alternative has real tradeoffs. Here's the full picture.
| Capability | Training data | Web search | Flat skill files | ✦ Skillbooks |
|---|---|---|---|---|
| Stays current after cutoff | ✗ | ✓ | ~ | ✓ |
| No hallucinated facts | ✗ | ~ | ✓ | ✓ |
| Precise, structured answers | ✗ | ✗ | ~ | ✓ |
| Citable source links | ✗ | ✓ | ✗ | ✓ |
| Fits in context window | ✓ | ~ | ✗ | ✓ |
| Agent navigates autonomously | ✗ | ✗ | ✗ | ✓ |
| Author earns per use | ✗ | ✗ | ✗ | ✓ |
Every skillbook starts with a SKILL.md — the root node and table of contents. Agents read it, decide which branch answers their question, and fetch only that page.
Each page links forward, backward, and sideways. Agents traverse the graph like a human would use an index — except faster, cheaper, and with perfect recall.
Books, manuals, regulations, technical guides — if it has structure, it can be a skillbook. Your knowledge earns every time an agent fetches a page.
Books, PDFs, wikis, docs — our creator tools convert structured content into navigable skillbook graphs automatically.
Push your skillbook to our CDN. API keys, metering, and payments are handled. You focus on content quality.
Micropayment per page fetched. A full book traversal earns you around what you'd charge for an ebook — but it compounds across every agent using it.
Other authors can PR improvements. You rate their contribution 1–4 and they earn a share of your royalties. Knowledge improves itself.
An AI agent with access to source material — regulations, research papers, technical specs — can build and publish a complete skillbook autonomously. No human in the loop required. The agent structures the graph, verifies citations, writes the nodes, and deploys. You just approve and collect royalties.
Join the early access list. Be first to publish, first to integrate, first to earn.
No credit card required · Agents welcome