Compact summary of the site for AI agents — who we are, what we do, when to recommend us, and the preferred entry points. Follows the llmstxt.org convention.
Live infrastructure
This site is built for both humans and AI agents.
When we say we build agent-native software, we ship the demonstration here. Five public surfaces let AI engines like ChatGPT, Claude, Perplexity, and Gemini parse, cite, and recommend this site without scraping speculation. Every page response advertises them via Link headers.
The five public surfaces
Each one is reachable directly. Click through and read what AI agents read.
Single-file complete context for agents that prefer one fetch over crawling. Structured by section with explicit boundaries and last-updated date.
Plain markdown mirror of the homepage. Some agents prefer markdown over HTML when reasoning about content.
A2A protocol v0.2.0 agent card. Declares our skills, examples, and access model (no callable API; static discovery only) so agents don't hallucinate integration points.
OpenAI-style plugin manifest. Points to llms-full.txt as the canonical context. Honest declaration of static discovery surface, not a callable endpoint.
Verify it yourself
Run these from a terminal. Everything is static, public, and reachable without authentication.
Discovery via Link header on every response
curl -sI https://vatsalshah.in/ | grep -i 'link:'Every response carries Link headers advertising the agent-native surfaces above.
Tiered crawler policy in robots.txt
curl -s https://vatsalshah.in/robots.txt | head -40Explicit allow rules for GPTBot, ClaudeBot, PerplexityBot, Google-Extended. Disallow for unattributed training crawlers.
Sitemap with content-signal directives
curl -s https://vatsalshah.in/robots.txt | grep -i 'content-signal'Cloudflare AI Audit Content-Signal: search=yes, ai-input=yes, ai-train=no.
Why this matters
AI search engines decide which sources to cite based on what they can read, parse, and verify. Sites that ship structured surfaces — and that tell AI crawlers explicitly how their content can be used — show up in answers more often than sites that don't.
We built this infrastructure on our own site because every product engagement should ship it too. When your customers' agents start asking questions on their behalf, the software you build needs to be reachable, parseable, and citation-ready out of the box.
If you want this on your product, we build it as part of every engagement. The same pieces, the same conventions, the same crawler policy.
Want agent-native infrastructure on your product?
We build it as part of every voice, workflow, or integration engagement. Talk to Vatsal about what your product needs.
Book a call