简体中文
EnglishWhat Is Lodd and What Does Its Web Analytics MCP Server Do?
Updated
Lodd is a web analytics MCP server built for AI agents. Instead of sending you to a dashboard, it exposes 42 MCP tools covering traffic, events, funnels, conversion attribution, and engagement metrics, so an agent like Claude Code, Claude Desktop, Cursor, or Codex can query your analytics inside the same conversation where you're already working. It's free up to 2,500 events per month, and setup is described as one prompt.
The core idea: analytics where the agent already is
Traditional analytics tools assume a human opens a browser, navigates a dashboard, and reads charts. Lodd assumes the opposite: your agent asks the questions and pulls the numbers directly.
The site frames this as replacing Google Analytics with "analytics that live where you work." The agent can check traffic, spot issues, and tell you what's working — and when you need something visual, it builds charts and tables on the fly rather than making you find a pre-built report.
What the 42 tools actually cover
The tool set spans five areas named on the site:
| Area | What the agent can pull |
|---|---|
| Traffic | Volume and trends over time |
| Events | Individual tracked actions |
| Funnels | Step-by-step drop-off between pages or actions |
| Conversion attribution | Which sources drove signups or conversions |
| Engagement metrics | How visitors behave once they arrive |
Because all of these live in one MCP server, a single question can combine them. The site's own example is "Why did signups drop?" — a question that needs traffic data, funnel analysis, and source attribution at once. A dashboard can't answer it in one pass; an agent with all 42 tools in the same conversation can.
What that looks like in practice
The site gives three concrete scenarios:
- Claude Code: the agent spotted a 34% traffic drop, traced it to a CSS regression hiding the pricing CTA, and shipped the fix — all without leaving the editor.
- Claude Desktop: asked for a conversion funnel, the agent built one from live data, found an 86% drop-off between pricing and signup, identified a 12-field form as the cause, and explained what to change.
- Claude Cowork: a scheduled weekly traffic check compared this week to last, flagged anything unusual, and delivered a summary before the workday started — catching a 45% drop in Reddit traffic before anyone noticed.
The same tools and data work across Cursor, Codex, and any other MCP client.
Who this is for
Lodd fits if you already work inside an MCP-capable agent and want analytics answers without switching context. The example dialogue on the site is telling: the agent didn't just report a 34% signup drop — it traced it to a Reddit post aging off the front page (67% of signup traffic), noticed the funnel broke from 28% to 11% between /pricing and /signup, and recommended two fixes.
That's the value proposition: not prettier charts, but an agent that can reason across traffic, funnels, and sources in one conversation.
It's less useful if you want a standalone dashboard for a team that doesn't use AI agents, or if you need analytics your whole org can browse without an MCP client.
Setup and pricing
- Setup: described as "one prompt to set up" — you copy an install prompt into your MCP client.
- Querying: "any prompt to query" — you ask in natural language.
- Free tier: up to 2,500 events per month.
- Privacy: the site lists "Private by design" as a heading, though the excerpt doesn't detail the specifics.
The site mentions pricing and upgrade as concepts but the provided material doesn't include actual paid tiers or prices, so check the site directly for current plans.
The short version
Lodd turns web analytics into something your AI agent can act on directly: 42 MCP tools for traffic, events, funnels, attribution, and engagement, queryable in natural language from Claude Desktop, Claude Code, Cursor, Codex, or any MCP client. Free up to 2,500 events/month. If your workflow already lives in an agent, it removes the dashboard step entirely.