The same data, machine-side.
Every section is a node. Statutes, regulations, cases and cross-references are typed edges. Query it over MCP, REST, or a single llms.txt — every response pinned to mpep-e9r01-2024-html for reproducibility.
claude mcp add mpep-remote \ --transport http \ https://api.mpep.io/mcp \ --header "Authorization: Bearer <key>"
curl -H "Authorization: Bearer <key>" \ "https://api.mpep.io/v1/search?q=obviousness" # → sections, cases, # cross-refs as edges
curl https://mpep.io/llms.txt # 736 sections, # one download
Tools over MCP
The mpep server exposes the corpus as typed tools — the same retrieval behind this site’s search.
searchBM25 full-text search across every sectionget_sectionFull section text with its cited-authority digestresolve_citation“MPEP § 2143.01”, “35 U.S.C. 112”, “KSR” → target + where the manual discusses itwhat_citesReverse lookup: every section citing a case or statuteread_caseFull opinion text for cited cases (CAP / CourtListener)get_form_paragraphExaminer form paragraph text by numbercoverageEdition, revision and structured coverage breakdownlist_chapterSection listing for a chapter with quality flagsTypical flow: search → get_section on the hits; a citation in hand → resolve_citation; “where does the manual discuss this case?” → what_cites.
License and access
The source texts (MPEP, USC, CFR, court opinions) are public domain. The structured data — citation graphs, annotations, holding summaries and compiled artifacts — is proprietary to Patent Agentic LLC (terms of use). Browsing is free; programmatic and bulk access beyond personal evaluation requires a key. This is not an official USPTO site. Made by the team at patentreply.ai.