Full catalogue power.
Zero catalogue exposure.
Your authors' methodology becomes interactive AI experiences your readers actually use — running on the complete book, with the source never leaving our servers.
You probably don't share full book text with AI platforms.
For good reason. Every existing AI tool — content licensing deals, training data partnerships, even retrieval-augmented chatbots — eventually exposes the source text. Once it's out, your authors' IP is gone. Margins compress. Catalogues commoditize.
So the AI experiences readers want — the ones that actually apply your authors' methodology to their real work — don't get built. The depth isn't there because the source can't be there.
We changed the architecture. Skills run server-side. Source files never leave our infrastructure. Your authors' complete books power the retrieval — but the books themselves can't be downloaded, scraped, or training-data appropriated.
The result: AI experiences that quote actual case studies, apply actual scoring rubrics, recognize actual edge cases — built from the complete work, not the public excerpts.
Your catalogue, fully working. Without the catalogue exposure.
Co-create with your authors. Earn from existing IP.
Start with 3-5 pilot titles.
We don't license your catalogue wholesale. We co-create with your authors, one title at a time, on a structure that aligns everyone:
- You bring the catalogue — methodology-rich titles where the framework is explicit and named
- Your authors bring approval — they retain control over how their methodology is presented and refined
- We bring the infrastructure — extraction pipeline (Director-led, conversational, no technical work for the author), empirical testing, server-side delivery, multi-platform distribution
- Readers bring their actual problems — the methodology runs on their work, not just explains itself
A pilot is typically 3-5 authors. We extract their methodology through guided conversation, build the skills, test them empirically, and ship them as a curated collection. Your authors earn ongoing revenue. You take a share. The catalogue extends into a new format that backlists never reach.
From your backlist to a working AI skill — in weeks, not years.
01ExtractProduction Mode — automated baseline.
Our AI extraction pipeline reads the author's complete work — book, supporting materials, prescribed templates and scoring rubrics — and produces a structured methodology draft. Calibrated thresholds. Decision trees. Prescribed formats. Pattern recognition. No author time required at this stage.
02RefineAuthor × Studio Pro.
The author reviews and refines the draft through guided conversation in Studio Pro. They sharpen the voice, adjust thresholds to match their actual standards, add nuance that didn't make the book. One to two sessions, typically 30-60 minutes each. Their methodology, in their voice, signed off by them.
03Test & ShipSix stages of empirical testing.
Methodology fidelity. Voice authenticity. Actionability. Differentiation against base AI. Scores published. Skill goes live on Claude via MCP — and built on the open Agent Skills + MCP standards, so it extends to ChatGPT, Cursor, Gemini, and VS Code as their MCP UX matures.
04EarnRecurring revenue. Real telemetry.
Authors earn 70% of every subscription. Publisher takes a partnership share negotiated upfront. Real-time analytics on usage, subscriber demographics, fidelity scores. The backlist becomes recurring revenue.
Why your full catalogue can power the skills — and stay safe.
Three architectural commitments that compound:
01Thin-client execution
Skills run on our servers. Methodology code, retrieval logic, response generation — all server-side. The end user's AI client only sees the output. Never the source.
02Full-text RAG, isolated retrieval
Your authors' complete books are indexed for retrieval — but the indexed content is encrypted at rest, accessible only to the skill's runtime, never returned to clients.
03DRM by architecture
Skills can't be copied because there's nothing to copy. No downloadable file, no extractable prompt, no exposed system context. The 'asset' is the running service, not a file.
Net effect: you can give us your full catalogue knowing the source stays where it belongs. Your authors get deeper skills. You stay protected.
What changes when your backlist becomes a service.
01New revenue from existing IP
Your backlist titles already exist. The methodology is already documented. Skills extend the asset into a recurring-revenue format without asking your authors to write new material.
02Stronger author-reader relationships
Readers don't just read the book once. They use the methodology in their actual work, repeatedly, year after year. Your authors become useful in an ongoing way — not just present in a Kindle library.
03Differentiation in a content-flooded market
The AI content flood is making books harder to discover and easier to commoditize. Interactive author skills aren't commodity content — they're a category your competitors can't ship without our architecture.
Built for publishers where every book contains a method.
Skills work best where books contain explicit, named methodology — frameworks with steps, structures with calibrated criteria, decision trees with prescribed formats. The kind of content that teaches a method rather than just describes a topic.
That points to a particular kind of publisher:
- Founder-led houses — where decisions can move fast and editorial control isn't fragmented across committees
- Methodology-density catalogues — business, professional development, leadership, sales, product, marketing, design, software, finance, learning
- Author-aligned economics — hybrid models or traditional houses that already give authors meaningful control over derivative formats
- Digital-first instincts — publishers already experimenting with online courses, interactive books, podcasts, communities
If that's you, we should talk. We're starting with 3-5 partner publishers in the first cohort. Pilot deals are 3-5 authors per house. We're not trying to be everything to everyone — we're building deeply with a small group of publishers who see this category coming.
Common questions from publishers we've spoken to.
- 01Who owns the IP after extraction?
- The author. We license delivery, not source. The methodology can be withdrawn from the platform at any time.
- 02How does revenue split work?
- Authors get 70% of every subscription. The remaining 30% covers infrastructure, distribution, and your partnership share — exact split negotiated with the publisher in the partnership agreement.
- 03What if our authors don't want to participate?
- Author participation is opt-in. We work directly with each author on their own methodology. No book becomes a skill without the author's explicit involvement and sign-off.
- 04Can skills be sold outside agent-skills.ai?
- Yes — skills can be distributed through the agent-skills marketplace, your own website, bundled with books, included in courses, or licensed to enterprise customers. Multiple distribution channels per skill.
- 05How long does a pilot take?
- Typically 6-10 weeks from partnership agreement to first skills live. Production Mode extraction is fast (days). The bottleneck is author refinement (one to two sessions per author) and testing (~1 week).
- 06What does the partnership cost?
- No upfront cost. Revenue share only. We invest in the extraction and infrastructure; you provide catalogue access and author coordination.
We're starting with 3-5 partner publishers.
If your catalogue contains methodology that gets results, let's talk. Initial conversation is 30 minutes — we'll know quickly if your catalogue is a fit.