jDocmunch MCP

Most agents still burn thousands of tokens dumping entire documentation files into context to find one config block. This server indexes docs once by section hierarchy, then lets you search and retrieve at heading granularity with byte-precise extraction. It parses Markdown, reStructuredText, AsciiDoc, Jupyter notebooks, HTML, YAML, JSON, and OpenAPI specs into stable section IDs that survive re-indexing. Tools cover discovery, structure browsing, semantic search, and single or batch retrieval. Sections include summaries, tags, content hashes, and parent/child wiring. Implements jMRI-Full with local-first storage under `~/.doc-index/`. Built for workflows where you need agents to navigate technical docs structurally instead of grep-and-pray or brute-force file reading.