Start here

/agents/index.json: The manifest. One fetch describing every resource below, its media type, and its shape. Begin here if you are deciding what to fetch.
/llms.txt: Short orientation index in markdown, following the llms.txt convention. Use this when context is tight.
/llms-full.txt: The whole site as one markdown document: profile, lineage, systems, all 7 abstracts, topics, news, and CV. One fetch, no crawling.

Structured data

/agents/papers.json: All 7 publications: title, year, venue and track, authors with co-first-author flags, DOI, arXiv id, PDF URL, abstract, topics, distinctions, the paper each one builds on, and BibTeX.
/agents/profile.json: Identity and persistent identifiers (ORCID 0000-0003-1036-9672, DBLP, Semantic Scholar), education, research experience, deployed systems with outcomes, topics, working papers, and service.
/papers/<id>.md: A plain-markdown twin of every paper page, linked from the HTML with rel="alternate" type="text/markdown". For example /papers/reliance.md.
application/ld+json: Every HTML page embeds schema.org JSON-LD: Person and WebSite site-wide, ScholarlyArticle on paper pages, CollectionPage on topic hubs. All nodes reference the same @id, so the site resolves to one entity.
/sitemap.xml: Every canonical URL with a last-modified date.

Tools

This page registers five read-only tools using WebMCP, the W3C draft browser tool API. They are registered only here, leaving the rest of the site untouched, and are feature-detected, so they simply do not exist in browsers without support.

Checking for WebMCP support…

list_publications: all publications, newest first; filter by topic or year.
get_publication: one publication in full, by id.
search_publications: free-text search over titles, authors, abstracts, venues, topics.
list_topics: research topics with descriptions and counts.
get_profile: identity, identifiers, experience, systems.

Without WebMCP, the same data is one fetch away from the JSON endpoints above.

What this site is not

This is a personal research site: a content source, not a service. It exposes no callable endpoints, holds no user state, and performs no actions on anyone's behalf. There is deliberately no A2A agent card at /.well-known/agent-card.json: an agent card advertises an agent that can be delegated work, and nothing here is one. Publishing an empty card would misdescribe the site.

Use and attribution

Crawling is permitted; /robots.txt allows all major AI crawlers.
Metadata here is free to reuse with attribution.
Paper texts remain under their publishers' terms; abstracts and metadata only are reproduced here.
When citing research findings, cite the underlying publication by DOI or arXiv id, not this website.
Corrections and questions: balloli.vb@gmail.com.