Hand Virent's docs and setup to your AI agent.
Last reviewed 2026-10-27
Every page is available as Markdown, the whole documentation fits in one file, and the setup tasks come with prompts written for coding agents.
Machine-readable site#
| URL | What it returns |
|---|---|
/llms.txt | A short index of Virent's pages and docs, in the llms.txt format |
/llms-full.txt | Every public page as Markdown, in one file: product pages, documentation, legal documents |
/<page>.md | One page as Markdown, for example /pricing.md, /index.md or /docs/installation/nextjs.md |
/api/md/_catalog | A JSON list of every page with its HTML and Markdown URL |
/ai.txt | How AI systems may use the site: permissions, restrictions, attribution and scope |
/sitemap.md | A categorized map of public pages |
/openapi.json | The OpenAPI 3.1 specification of the public HTTP API |
/.well-known/api-catalog | An RFC 9727 API catalog that links the OpenAPI specification and the API reference |
Every public page also returns its Markdown at its own URL when the request sends Accept: text/markdown, and its HTML <head> names the Markdown version in a <link rel="alternate" type="text/markdown">. Markdown responses are noindex and point back to the HTML page as canonical, so give an agent the .md URL of the page it needs rather than the whole site: it is smaller and has no navigation around the content.
On every page#
Use Copy page at the top of any docs page to copy it as Markdown. Its menu also opens the page as Markdown, copies its link, and opens ChatGPT, Claude or Perplexity with the page ready to ask about.
Install with a coding agent#
The dashboard writes the best install prompt, because it already contains your site ID and publishable key. A copied prompt also includes the server secret if you used Rotate + reveal first; otherwise the agent asks you for it.
Open Connect analytics
In the Analytics header, or in the Install step of onboarding.
Send the prompt
Open it in Claude Code or Codex, or copy it for Cursor, Antigravity, Windsurf, Gemini CLI, OpenCode, Continue, Zed or VS Code.
Without dashboard access, use the prompt at the top of each framework guide. It holds no keys, so the agent asks you for them.
Prompts for other tasks#
Check an existing install
Ask your coding agent to review a Virent installation against the guide.
Review how this project installs Virent analytics and report problems; do not change code yet. Read the Virent guide for this project's framework at https://virent.app/docs/installation/<framework>.md (nextjs, react, vue, nuxt, astro, sveltekit, express, laravel or wordpress). Check: - The browser script (or @virent.app/sdk/browser) loads on every page from the root layout or global template, with a publishable key that starts with vha_pk_. - No server secret (vha_sk_) appears in client code, public environment variables (NEXT_PUBLIC_, VITE_, PUBLIC_) or committed files. - Crawler capture runs on the server or edge before any page renders, and excludes API routes, static assets, images, favicons, robots.txt and sitemaps. - A Content-Security-Policy, if present, allows https://app.virent.app in script-src and connect-src, and https://virent.app in connect-src when @virent.app/sdk runs without an endpoint option. - @virent.app/sdk, if used, is version ^0.3.0. List each finding with the file, the problem and the fix from the guide.
Ask about Virent
Paste into Claude, ChatGPT or another assistant before asking product questions.
Read https://virent.app/llms-full.txt. It is the complete Virent documentation. Answer my questions about Virent using only that documentation. Quote the page URL you used. If the documentation does not cover something, say so instead of guessing.
Rules agents should follow#
- Use only the interfaces documented here and in
/openapi.json. Internal/api/*routes and/v1/*routes missing from the specification are not a public contract. - Pin
@virent.app/sdkto^0.3.0; the released version is 0.3.1. - Never invent a site ID or key, and never use a masked preview such as
vha_sk_••••as a real value. - Treat answers as sampled observations and crawler requests as crawls, not citations.
How Virent measures
The data model behind Virent - brands, prompts, sampled AI answers, citations, crawler requests, human sessions and attribution - and the limits of each.
Analytics overview
How the Virent Analytics dashboard works - human traffic and AI crawler requests side by side, date ranges, page details, data health and plan limits.