nextanalytica: headless web analytics for your own AI agent
Everything nextanalytica offers, in four layers: where you use it, what it does, the data it runs on, and the rules it keeps.
Updated
nextanalytica is headless web analytics for your own AI agent: one tag collects your site's visits, your agent (Claude, Cursor, Codex or any MCP client) reads them through our MCP server, and skills on a schedule tell your team what changed and what to fix. Everything it offers is one of four layers.
Surfaces: where you use it
Each has its own job. None of them needs the others, except the tag.
- Your agent, over MCP: Claude Code, the Claude app, Cursor, Codex or any MCP client reads the numbers, answers in chat, and acts only through the tools your team connected.
- The skills plugin for Claude Code: your project's skills as slash commands, kept in step with the Skills window.
- The web app: add a site and its tag, choose who has access, connect sources, switch skills on. Everyone on the team sees the same, with or without an agent.
- Email: when a scheduled skill finds something, its owner gets a short note with no data in it; the finding waits in
get_context.
Capabilities: what it does
- Web analytics: visits, pages, referrers and UTM tags, countries, devices and speed (Web Vitals), by day. AI assistants (ChatGPT, Perplexity, Claude and others) are their own channel.
- Product analytics: people (after the visitor's consent), funnels, retention, paths and cohorts, from one
querytool. - Conversions: what counts as a win, said once (a page, an event, a form or an order) and counted back over the days we keep.
- Sources: Google Search Console, Google Analytics 4, Bing Webmaster Tools, ChatGPT Ads, Odoo, copied read only next to the visits they bring.
- Skills: instructions your agent runs on a schedule, built in (Weekly check-up, Wasted ad spend, Landing page fixes, Form drop-off, Traffic drop detective, Search titles, Build my dashboard) or your own. They run on your own Claude Code, never on our AI.
- Connectors: the tools your team connected (Linear, HubSpot or any remote MCP server). Your agent acts only through them, and every call is logged.
- Project knowledge: the team's instructions and memory (goals, customers, decisions), shared by every person and every agent on the project.
- Timeline: dated launches, campaigns, price changes and outages, so "why did it drop on Tuesday" has an answer.
Data: what it runs on
One tag on the site sends page views (single-page navigation included), outbound clicks, downloads, forms, errors, Web Vitals and the site's own events. Until a visitor agrees, it counts without cookies and keeps nothing across visits. The sources' copies sit next to it. Everything is kept 30 days.
Trust: the rules it keeps
- Each project's data is kept apart; one project can never read another's.
- Each person signs in with their own account; there are no shared keys.
- We run no model on your data. The thinking happens on your own agent.
- Agents change your site or tools only through connectors an owner or admin added.
Connect an agent: MCP brief. Install the tag: install. Questions: FAQ. Prices: pricing. Every page for agents: llms.txt.