Glossary
Model Context Protocol (MCP)
An open standard for connecting AI applications to external systems. An analytics tool with an MCP server lets an assistant read its figures.
The Model Context Protocol is an open standard for connecting AI applications such as Claude or ChatGPT to external systems. A system exposes tools through an MCP server, and an assistant can then call them. Its specification defines two transports: stdio, for local servers, and Streamable HTTP, for remote ones.
For analytics, it means you can ask an assistant where people drop off in a signup flow, and it reads the answer from the tool itself instead of a pasted screenshot.
Trailic’s MCP server is at /api/mcp. A member signs in and approves the assistant, and each tool runs with that member’s rules. Nearly all of them only read: site_overview, top_pages, run_funnel, run_retention and list_events, for example. See the features page. Other tools have MCP servers too, including PostHog, Umami and Vercel (checked November 2026).