Agentlytics Turns AI Agents Into Marketing Operators

Overview

A new tool called agentlytics is attempting to bridge the gap between web analytics and autonomous AI agents, transforming raw traffic data into actionable marketing strategies. Built on the Model Context Protocol (MCP), the service allows coding assistants like Claude Code, Cursor, and Codex to not only read website metrics but also interpret them and execute growth campaigns.

The core value proposition is simplicity: users add a single JavaScript snippet to their site and then prompt their AI agent to analyze performance. The agent can then identify traffic spikes, explain their origins, and generate tracked marketing links to replicate success. By removing the need for manual dashboard analysis, agentlytics aims to make analytics a direct input for development and marketing workflows.

"Analytics your AI agent can read, explain and act on. Add one script tag. See every traffic spike, where it came from and why." — agentlytics

How It Works

Agentlytics operates by feeding structured data directly to AI agents via MCP tools. Once a lightweight, cookieless tracking script is embedded in a website’s header, the system begins recording pageviews, referrers, UTM tags, and custom events such as signups or purchases. This data is then accessible to connected coding agents, which can query it using natural language.

The platform emphasizes spike detection as a primary feature. Rather than presenting static charts, agentlytics identifies unusual jumps in traffic by comparing current hourly data against the same hour in previous weeks. This contextual comparison helps distinguish genuine growth events from regular daily fluctuations. When a spike is detected, the agent provides an AI-generated summary that ranks the causes by impact, highlighting the specific referrer, landing page, and conversion events responsible for the surge.

For example, if a Show HN post drives a 300% increase in traffic with a high signup rate, the agent can immediately explain this correlation. It can then suggest follow-up actions, such as drafting a secondary post with tracked links to target the same audience. This creates a closed loop where data observation directly informs content creation and deployment.

The marketing loop is designed to be autonomous. Agents can plan weekly marketing efforts based on historical performance data, create unique tracked links for each campaign, and report back on results. This allows developers to treat marketing as another variable in their codebase, managed alongside feature development.

Key Features and Capabilities

The service offers a focused suite of MCP tools that grant agents precise control over analytics and reporting. These tools enable agents to retrieve overviews of visitors and pageviews, find and explain traffic spikes, and query visits based on filters such as country, browser, or source.

Automated Insights and Growth Actions

One of the standout features is the ability to generate growth actions. Instead of vague advice, the system ranks suggestions based on real channel performance and conversion rates. If data shows that Reddit users convert at a significantly higher rate than other sources, the agent will prioritize drafting content for that platform.

Agents can also manage campaign tracking automatically. Every link published by the agent is tagged with UTM parameters, allowing for granular attribution. Users can then ask the agent to review campaign results, receiving clear metrics on visitor counts and conversion rates for specific initiatives.

Privacy and Compliance

Agentlytics markets itself as a privacy-first solution, appealing to websites concerned about regulatory compliance and user trust. The tracking script does not use cookies, meaning sites do not need to display consent banners for analytics purposes.

Furthermore, the system employs hash-based anonymization for visitor identification. IP addresses are never stored, and the daily-changing hashes ensure that individual users cannot be tracked across time. To prevent data leakage, bot traffic from crawlers and headless browsers is filtered out before it reaches the analytics engine. Each read key is scoped to a single verified domain, ensuring that data remains isolated and secure.

Setup and Integration

Onboarding is designed to be rapid, with the company claiming a setup time of approximately two minutes. The primary method involves pasting a single command into a compatible coding agent. For instance, users can prompt Claude Code, Cursor, or Codex to add agentlytics to their project by following instructions found at a specific URL.

The agent then guides the user through signing up with an email and domain, adds the tracking snippet to the site’s <head> section, and configures the MCP connection. For users who prefer manual setup, the service supports a wide range of platforms, including Next.js, React, Astro, WordPress, Webflow, and Shopify.

A public dashboard is available for transparency, allowing users to view live metrics without needing to log in. This dashboard displays key indicators such as total pageviews, visitor counts, and signup numbers, along with visual indicators for recent traffic anomalies.

Pricing and Availability

Agentlytics operates on a freemium model. The free tier includes up to 5,000 pageviews per month, making it accessible for personal projects, startups, and early-stage products. There are no credit card requirements to start, and all features are included across all plans.

Pricing scales with usage, allowing users to upgrade as their traffic grows. The company emphasizes that plan differences are limited to pageview limits and the number of sites supported, rather than gating core features behind paywalls. This structure encourages long-term adoption as projects mature and their analytics needs expand.