Analytics for AI agents.
Analytics for AI agents is the category where visitor data is exposed via CLI + MCP server so any LLM can call it like a native tool, no glue code. Eyepup defines this category. Your Claude Code, Cursor, Codex, or custom agent reads heat scores, friction patterns, and session evidence the same way it reads files — and emits the next fix.
Why this is a category, not a feature
Most analytics tools were built for humans clicking through dashboards. The interfaces — charts, filters, date pickers — are optimised for eyeballs, not for LLMs. To use them, an AI agent has to either screenshot the dashboard (lossy) or scrape an undocumented API (brittle).
Analytics for AI agents flips the default. The primary surface is text: a CLI you can pipe, a JSON API the agent can call, and an MCP server the agent calls without code. The dashboard exists for humans, but the agent is a first-class citizen — same data, same auth, structured output.