Comment by sabri_eker

I implemented a process. - First snapshot. Generated day by day files based on last 2years commit, used Claude to enrich every commit message (why,how,where and output). Also create File.java.md file enriched commit message per file. - I have the history. Next I embedded all of this to Postgresql database. - Implemented an MCP to query anything. - Create whatchdogs to followup projects that I setup, git hook creates stub files per commit. - And then process enriches new stub files. then indexes it to Postgresql.

I did for all internal projects. And first rule of CLAUDE.md is ask MCP for any information. Now Claude knows what other related projects did, what should adopt for.