[Feature] Add support for MCP server

[ahnopologetic]

Describe the feature

Add support for parsing and generating tracking documentation from MCP (Measurement Collection Protocol) server sources. MCP is often used in custom tracking infrastructure to collect analytics events via HTTP endpoints, particularly in data-driven environments where teams manage their own pipelines outside of major vendors like Segment or Amplitude.

Support for MCP server would involve:

• Recognizing and interpreting custom event payloads sent to an MCP endpoint
• Mapping event structures to a unified schema format
• Enabling CLI or API options to specify MCP log or endpoint inputs (e.g. .jsonl, .ndjson, or proxy logs)
• Optional integration with OpenAPI/Swagger if MCP endpoints are defined in API specs

This would expand the scope of @flisk/analyze-tracking beyond SDK-based clients to server-based data collection strategies.

Describe alternatives you've considered

Considering the possibility that you want to keep this project as a simple CLI, we can create a new repository to do so.
However, it seems we can expose some public functions that would act as tools, off of this codebase. (e.g., listEvents, getEvent, addEvent, etc.)

Who will this benefit?

Me, future product engineers with AI-enhanced environment, e.g., Cursor, Windsurf, and so forth

Are you interested in contributing this feature?

Yes

Anything else?

No response

[skarim]

Sorry, but this reads like an LLM hallucination. Can you please explain in plain language what the feature request is?

AFAIK "Measurement Collection Protocol" is not a thing. Google does have a measurement protocol that they have defined for use in Google Analytics. Did you mean that? If so, that would require integrating with Google's APIs to extract data from GA and turn into schemas.

MCP is often used in custom tracking infrastructure to collect analytics events via HTTP endpoints

Can you give some examples here? If by MCP you mean Model Context Protocol, those servers are used to connect LLMs with external tools/datasets. I am unaware of an LLM being embedded in a server to log analytics data via HTTP (and that sounds like a very expensive way of logging data).

The project currently does support server-based data tracking, by detecting function calls in the codebase.

[ahnopologetic]

Yes. It's my bad. I wrote a draft of this github issue and ran LLM to refine it. I meant MCP (model context protocol)

What I mean by that was:

1. Let AI-enhanced clients (e.g., cursor, windsurf, claude, etc) read what event tracking codes are existing: without needing to read through codebase, they can find exact
2. Make sure the output yaml/json file is in sync: Say we type in "Sync events" in cursor. It will look up predefined yaml/json file we created, look up the code location and double-check the code is present (e.g., "Cart Added" - src/some/component/card.tsx:16, is the code present? good, else? you add or edit it).
3. Simplified event code management: Potentially, we can manually or let AI edit the yaml/json file to add more events or update them, and let AI do the tedious event tracking code update works, declaratively.

Those are usecases I can think of on top of my head.

We can do this by exposing some of this repo's internal functions/methods like:

• listEvents
• getEvent
• addEvent
• updateEvent
• deleteEvent
  with proper documentation (jsdoc in our case).

Hope this clarifies it better.