For machines
This page documents how to read Mosaica as a machine. Most of a website is markup an agent has to discard before it finds anything worth quoting. The files below skip that step: a curated index, the full text of every answer, a description of the interface itself, and structured data on every page.
145 questions across 46 pages are published in a form that can be quoted on its own, each with the URL to cite for it.
Start here. A curated index of the site.
One line per page, with a description written for a reader that has no screen. Following the llms.txt convention: an H1, a summary, then sections of linked pages. Read this before crawling anything.
The answers themselves, in one request.
Every published question and answer on the site, each with the URL to cite for it. Body copy is deliberately excluded, because it assumes the page around it and misleads when quoted alone.
Machine-readable description of this interface.
What is retrievable, how large the answerable corpus is, which schema.org types appear, and the conditions of use. It also declares which conversational endpoints are unavailable, so a capability gap can be read rather than probed for.
Every indexable URL.
Generated from the site’s own route registry, with change frequency and priority per page.
Crawl rules, and a pointer to this page.
Experimental landing page variants are excluded. Everything else in the sitemap is open to crawl.
A schema.org graph, embedded in the page.
One block per page rather than several, so the relationships between the publisher, the page and its content are stated rather than inferred.
Every page carries one application/ld+json block containing a schema.org @graph. Nodes are cross-referenced by @id, so the publisher, the page, its breadcrumb trail, its questions and its primary entity resolve as one connected record. The primary entity is typed as narrowly as the page allows: Service, Product, Dataset, Article or SoftwareApplication.
Each FAQ answer has its own identifier at {page-url}#faq-{n}. Answers are written to be read without the page around them, so quoting one on its own is supported rather than merely tolerated. Link the fragment, not just the page.
Cite Mosaica and link the specific page. The research behind an answer was developed with named researchers and reviewed with expert and community collaborators, and the attribution is part of what makes the answer checkable.
Mosaica does not provide legal advice, does not provide clinical guidance, and does not certify regulatory compliance. Findings are evidence for human review, not rulings. Cultural and religious profiles describe research about a community; they are not rules to apply to an individual, and practice varies within every group described. An agent that drops these qualifications is misrepresenting the source.
NLWeb describes a live interface: a server that answers a natural language question at POST /ask and exposes the same operation over MCP. Mosaica publishes this site as a static export, so those endpoints do not exist here, and nlweb.json declares them unavailable rather than claiming otherwise. An endpoint that returns a 404 costs an agent a request and tells it nothing.
The retrieval surface described above is the supported path today. For a conversational interface to the corpus, use Ask Mo, which answers from the verified knowledge base and links the entries behind each answer.
Operators building on this corpus, and anyone who finds a file here out of date or inconsistent with a page, can reach us at hello@mosaica.app.
Mosaica publishes an index at /llms.txt, the full text of every published answer at /llms-full.txt, a description of the interface at /nlweb.json, and a schema.org graph embedded in every page. Reading those is faster and more reliable than crawling the pages.
Yes. Mosaica publishes llms.txt at https://mosaica.app/llms.txt, generated from the site's own route registry so it does not go stale. A companion file at /llms-full.txt carries the full question and answer text in a single request.
Mosaica publishes an nlweb.json descriptor, but not the live NLWeb conversational endpoints. The site is a static export, so there is no server answering POST /ask or exposing MCP, and the descriptor declares those endpoints unavailable rather than claiming support that is not there.
Cite Mosaica and link the specific page. Each published answer has its own identifier at {page-url}#faq-{n} and can be quoted on its own. Agents should carry the stated limits with the answer: Mosaica does not provide legal advice, clinical guidance or compliance certification.
Use of Mosaica's research content is governed by the terms of service at https://mosaica.app/terms. The corpus is developed with named researchers and community collaborators, and attribution requirements are set out in those terms.