Internet Movie Script Database - MCP for Movie Catch Up Tool

This project is an MCP to connect LLM's to the IMSDB database. This MCP is being built to serve a tool that will be able to tell the user a summary of a show so far, so they can catch up without any spoilers. This is the first version, which will require the user to give a summary of what is happening now. In the final verison, the user should be able to input the time stamp, then the LLM will be able to return that summary.

Workflow:

1. User gives movie title + a description of where they are

2. Check cache (SQLite) for that movie title

3. If not cached: scrape IMSDb → parse into scenes (your number-matching + all-caps + INT/EXT logic) → generate one-line summary per scene (LLM, done once) → insert all rows into SQLite

4. If cached (or now freshly cached): LLM matches user's description against the scene_summary column to find the right scene_number

5. Pull scene_text for all scenes 1 through N → LLM synthesizes a spoiler-free "here's what's happened" recap

Related MCP server: Movie Search MCP Server

Data model (SQLite, one table):

scenes: movie_title (TEXT), scene_number (INTEGER), scene_summary (TEXT), scene_text(TEXT)