{"apiVersion":"v1","directorySlug":"usa-mcp-mcp-4ed444d5cc","title":"usa·mcp","description":"A curated, provenance-tracked librarian over primary sources on United States history, served to your AI assistant over MCP. Every claim traces to a verbatim passage — and it says so when the record cannot answer.","pageUrl":"https://www.mcpbundles.com/skills/usa-mcp-mcp-4ed444d5cc","connectPath":"importable","connectAction":"import_server","catalogListing":false,"bundleSlug":null,"listingKind":"remote_mcp","kind":"provider","transportUrl":"https://usa-mcp.com/mcp","hostedMcpUrl":null,"toolCount":9,"toolInventoryStatus":"observed","refreshedAt":"2026-11-16T11:41:43.887283+00:00","registryName":"com.usa-mcp/usa-mcp-mcp-4ed444d5cc","registryVersion":"2026.11.16","registryEligible":true,"tools":[{"name":"orient","description":"START HERE. Returns the serving rules, the topic/facet map, `corpus_stats` (the authoritative counts — use these for any claim about the library's size or coverage, never an estimate) and `index_facets` (what passages() can filter on). It does NOT enumerate the sources: that list is fetched on demand via sources(), so the connect-time payload stays flat as the corpus grows. Read this first.","inputSchema":{"type":"object","properties":{},"additionalProperties":false},"logicalName":"orient","displayName":"orient","title":"orient","annotations":{"untrustedContentHint":true,"readOnlyHint":true}},{"name":"route_question","description":"Map a user question to the curated topic file and its sources, with the tiers and vantages in play, each source's stance and source criticism, and a caveat. Use this FIRST for any substantive question; search is only a fallback.","inputSchema":{"type":"object","required":["question"],"properties":{"question":{"type":"string"}},"additionalProperties":false},"logicalName":"route_question","displayName":"route_question","title":"route_question","annotations":{"untrustedContentHint":true,"readOnlyHint":true}},{"name":"read","description":"Read a whole committed knowledge/ or map/ file (e.g. a topic file). Layer 1 source texts are not readable whole — use cite for those.","inputSchema":{"type":"object","required":["path"],"properties":{"path":{"type":"string"}},"additionalProperties":false},"logicalName":"read","displayName":"read","title":"read","annotations":{"untrustedContentHint":true,"readOnlyHint":true}},{"name":"search","description":"Capped keyword search across source texts. EXPLICITLY NON-EXHAUSTIVE and allowed to miss (exhaustive=false); a no-hit result is not proof the library lacks an answer. Snippets are context, not citations — call cite before quoting. Prefer route_question.","inputSchema":{"type":"object","required":["query"],"properties":{"limit":{"type":"integer","maximum":25,"minimum":1},"query":{"type":"string"}},"additionalProperties":false},"logicalName":"search","displayName":"search","title":"search","annotations":{"untrustedContentHint":true,"readOnlyHint":true}},{"name":"cite","description":"Return the verbatim passage behind a claim, with full citation, the source's own section (chapter, or the individual informant's name — cite that, don't guess), the source criticism, and the upstream provenance chain — or an ERROR if the locator does not resolve. This is proof, not paraphrase. If it errors, do NOT invent the passage. PASTE `quotable` — it is the matched text with its citation already attached and is correct by construction. `context_not_for_quotation` is the surrounding window,","inputSchema":{"type":"object","required":["source_id","locator"],"properties":{"window":{"type":"integer","maximum":1500,"minimum":120},"locator":{"type":"string"},"source_id":{"type":"string"}},"additionalProperties":false},"logicalName":"cite","displayName":"cite","title":"cite","annotations":{"untrustedContentHint":true,"readOnlyHint":true}},{"name":"check_answer","description":"Resolve every quotation in a DRAFT answer against the corpus, before you send it. Deterministic and free — no model involved. Returns any quotation this corpus cannot confirm, with the closest actual text where there is one. QUOTATION MARKS ARE A CLAIM OF VERBATIM and this is how you check the claim: run it on your draft, fix anything under `unresolved`, then answer. It is exact-match only — it cannot tell you whether a resolving quotation is FRAMED honestly, which remains yours to get right.","inputSchema":{"type":"object","required":["draft"],"properties":{"draft":{"type":"string"}}},"logicalName":"check_answer","displayName":"check_answer","title":"check_answer","annotations":{"untrustedContentHint":true,"readOnlyHint":true}},{"name":"check_absence","description":"BEFORE you write that this corpus lacks something, check it here. Absence is the one claim you cannot support by citing a passage, so agents infer it from a search that came back empty — and search is capped, whole-word, and does not stem. This searches the Layer 1 text directly with aliases and stemming, and requires the terms to CO-OCCUR, not merely to appear in the same file. NOT ABSENT means do not claim the gap. SCATTERED means the words exist but not together — say the corpus does not hold","inputSchema":{"type":"object","required":["claim_terms"],"properties":{"claim_terms":{"type":"string"}}},"logicalName":"check_absence","displayName":"check_absence","title":"check_absence","annotations":{"untrustedContentHint":true,"readOnlyHint":true}},{"name":"sources","description":"List the library's sources — filter by topic, tier, or a single source_id. Per-source detail is deliberately NOT inlined in orient (it grew ~668 bytes per source and every byte is paid on every session), so fetch it here. Summary rows omit role/stance/source_criticism; ask for one source_id to get the full record including its provenance chain. For corpus SIZE use orient's corpus_stats, never a page count from here.","inputSchema":{"type":"object","properties":{"tier":{"type":"string"},"limit":{"type":"integer","maximum":100,"minimum":1},"topic":{"type":"string"},"offset":{"type":"integer","minimum":0},"source_id":{"type":"string"}},"additionalProperties":false},"logicalName":"sources","displayName":"sources","title":"sources","annotations":{"untrustedContentHint":true,"readOnlyHint":true}},{"name":"passages","description":"Filter the passage index by THEME (what it's about), ENTITY (who/where — a person, place or case) and/or DATE. Use theme and entity as entry points; date is a refinement, not a way in. Returns POINTERS, not claims: each result says only that this text exists at this location — call cite() to read it. NOT exhaustive; absence here is not evidence of absence, and you must not count results and report the number as how common something was. If a curated theme file exists for the subject, read it fir","inputSchema":{"type":"object","properties":{"limit":{"type":"integer","maximum":50,"minimum":1},"theme":{"type":"string"},"entity":{"type":"string"},"offset":{"type":"integer","minimum":0},"date_to":{"type":"string"},"date_from":{"type":"string"},"source_id":{"type":"string"}},"additionalProperties":false},"logicalName":"passages","displayName":"passages","title":"passages","annotations":{"untrustedContentHint":true,"readOnlyHint":true}}],"toolSource":"directory_preview","agentAction":"both","installLinks":[{"id":"cursor","label":"Cursor","href":"cursor://anysphere.cursor-deeplink/mcp/install?name=usa-mcp&config=eyJ1cmwiOiAiaHR0cHM6Ly91c2EtbWNwLmNvbS9tY3AifQ%3D%3D"},{"id":"vscode","label":"VS Code","href":"vscode:mcp/install?%7B%22name%22%3A%22usa-mcp%22%2C%22type%22%3A%22http%22%2C%22url%22%3A%22https%3A%2F%2Fusa-mcp.com%2Fmcp%22%7D"},{"id":"vscode-insiders","label":"VS Code Insiders","href":"vscode-insiders:mcp/install?%7B%22name%22%3A%22usa-mcp%22%2C%22type%22%3A%22http%22%2C%22url%22%3A%22https%3A%2F%2Fusa-mcp.com%2Fmcp%22%7D"},{"id":"claude-web","label":"Claude.ai","href":"https://claude.ai/customize/connectors?modal=add-custom-connector&connectorName=usa%C2%B7mcp&connectorUrl=https%3A%2F%2Fusa-mcp.com%2Fmcp"},{"id":"codex","label":"Codex","copyText":"codex mcp add \"usa-mcp\" --url \"https://usa-mcp.com/mcp\""},{"id":"chatgpt","label":"ChatGPT","href":"https://www.mcpbundles.com/docs/integrations/chatgpt"}],"catalogUrl":"https://www.mcpbundles.com/api/directory/v1/servers/usa-mcp-mcp-4ed444d5cc","callUrl":"https://www.mcpbundles.com/api/directory/v1/servers/usa-mcp-mcp-4ed444d5cc/tools/{tool}","mcpServerUrl":"https://usa-mcp.com/mcp","liveCatalogUrl":"https://www.mcpbundles.com/api/directory/v1/servers/usa-mcp-mcp-4ed444d5cc?live=1"}