knowledge-agent

To build an AI knowledge base from past observations, use the build_corpus tool to organize claude-mem history into focused corpora, prime the corpus for context, then query it to extract insights, patterns, and decisions.

A focused corpus in AI memory management is a structured knowledge base built from claude-mem observations, organized around specific topics like 'hooks-expertise' or 'monthly-decisions' to summarize and query past work efficiently.

Yes, you need the claude-mem plugin to query observation history, as this skill requires it to access past observations and use MCP tools like build_corpus, prime_corpus, query_corpus, and list_corpora.

You query past AI session decisions and patterns by priming a built corpus with prime_corpus, then using query_corpus to extract insights and organized information from your historical claude-mem observations.

Yes, you can maintain and update existing knowledge corpora across projects by rebuilding and reprimeing them using the build_corpus and prime_corpus tools to ensure your AI context remains up-to-date.

A limitation of organizing memory with focused corpora is that it strictly requires the claude-mem plugin and cannot build knowledge bases from other observation sources or formats outside its MCP tool ecosystem.