# Desvela

> The watchman of the agent economy. Desvela watches the web with cheap deterministic checks and wakes your agent — or your team — with a signed, structured diff. Your agent sleeps. Desvela does not.

Desvela sells watches, not dashboards. First product: AI Brand Watch — track a brand's mention rate in ChatGPT, Claude, Perplexity and Gemini answers, pay per check, delivered as HMAC-signed webhook events. Zero-config: just a brand name — the category questions, competitors and domain are discovered and frozen as a prompt set, so every scheduled run measures the same questions and the diff is comparable. MCP access is live via Apify's gateway (https://mcp.apify.com/?tools=desvela/brand-watch, Apify token as Bearer). x402 agentic payments are live too: an agent can run Brand Watch paying USDC on Base with no Apify account, billing or API key (POST https://api.apify.com/v2/acts/desvela~brand-watch/runs without credentials returns 402 with the payment terms). Registry Watch is live too (beta, free): subscribe to any domain and get a signed webhook when its ai-catalog.json, llms.txt, agents.md or robots.txt AI directives change. Roadmap: Orient (wake-up briefing for agents), native MCP server under the ai.desvela/* namespace.

## Pages

- [Home](https://desvela.ai/index.html): positioning, the three moments of the agent lifecycle (Orient, Preflight, Watch & Wake), process and sample change event
- [AI Brand Watch](https://desvela.ai/brand-watch): product page — zero-config (just a brand name, questions discovered and frozen as a prompt set), how it works, mention-rate tracking, competitive share of voice, owned vs earned citations, pricing ($0.02 actor start, $0.08 per query×engine, $0.04 prompt discovery, $0.04 GEO recommendations, $0.04 sentiment, alerts free, ≈$14/month typical), honest measurement FAQ
- [How it works](https://desvela.ai/how-it-works): the pipeline — watch, sample, diff, wake — and the token-economics argument
- [For agents](https://desvela.ai/for-agents): webhook payload contract, delivery modes (signed webhook push, dataset pull, or Slack/email/Zapier via Apify integrations), bot policy, live MCP endpoint, x402 agentic payments
- [Registry Watch](https://desvela.dev/registry-watch): subscribe to a domain, signed webhook when its agent surfaces change, weekly checks, free during beta
- [AI Discovery & Access Signals Checker](https://desvela.ai/check): inspect the public files and policies that AI agents and crawlers may use to discover or access a site, graded against the census baseline. It does not measure AI citations or recommendations
- [The census](https://desvela.ai/census/): monthly State of the agent-readable web, Tranco top-100K
- [Why the numbers undercount](https://desvela.ai/measurement): what an AI-visibility number can and cannot measure — three identical runs six minutes apart, per-engine source stability (Perplexity 96%, ChatGPT 9%), and why one sample per prompt is a coin flip
- [Memory or live search](https://desvela.ai/memory-or-search): the OpenAI API answers from memory unless the search is forced — left alone it searched for 1 question in 5, and forced vs unforced runs agreed on only 60% of the recommended brand set
- [Docs](https://desvela.dev): quickstart, input schema, pricing events, webhook payload reference, HMAC signature verification, retries

## Try it

- [AI Brand Watch on Apify Store](https://apify.com/desvela/brand-watch): runs on Apify, pay per check, schedule weekly

## Contact

- Email: hello@desvela.ai
- X/Twitter: https://x.com/desvela_ai
- Bot policy: https://desvela.dev/bot
- ARD manifest: https://desvela.ai/.well-known/ai-catalog.json (this domain) and https://desvela.dev/.well-known/ai-catalog.json
- The census (monthly, State of the agent-readable web): https://desvela.ai/census/
- ARD registry search (POST /search, ARD spec): https://registry.desvela.dev/search
- Free preflight(domain) MCP tool: POST https://registry.desvela.dev/mcp (no auth)
- Free crawler_block_test(domain) MCP tool, same endpoint: what a site's CDN actually serves GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and CCBot, versus what its robots.txt declares. Also GET https://registry.desvela.dev/crawler-block?domain=