# Machine Library > Machine Library is the search and AI product by Space Frontiers Company. It is a full-text retrieval API and MCP server over millions of academic papers, books, patents, Wikipedia articles, YouTube transcripts, and live social posts from Reddit, Telegram, and Discord. Built for AI agents doing literature review, fact-checking, citation walking, and grounded research synthesis. It provides compact search snippets plus bounded full-text retrieval where source content is available. Two corpora, one API: - **documents** — peer-reviewed papers (CrossRef, PubMed, arXiv), books, patents (USPTO), Wikipedia, technical standards, manuals, and YouTube video transcripts. - **social** — Reddit, Telegram, and Discord. Three integration surfaces: - **MCP server** — `https://mcp.machinelibrary.ai` (Streamable HTTP transport, OAuth 2.1 with PKCE or Bearer API key). Drop-in for Claude Desktop, Claude Code, Cursor, Windsurf, Cline, and any MCP-compatible agent. - **REST API** — `https://api.machinelibrary.ai` (OpenAPI 3.1). Core endpoints include `POST /v2/search/`, `POST /v2/search/similar`, token-bounded text fetches at `GET /v2/documents/{document_id}` and `GET /v2/documents/by-uri/{uri}`, original-file streams at `GET /v2/downloads/{document_id}.{format}` and `GET /v2/downloads/by-uri/{uri}`, and `POST /v2/conversations/`. Full machine-readable spec at `https://machinelibrary.ai/docs/openapi.json`. - **Web app** — `https://machinelibrary.ai/search` for search and cited AI answers in the browser; saved conversations live at `https://machinelibrary.ai/c/{slug}`. Core read-only MCP retrieval tools: - `machinelibrary_search_documents` — search papers, books, patents, Wikipedia, and YouTube transcripts. Use for peer-reviewed claims, citations, prior art, and video sources. - `machinelibrary_search_social` — search Reddit, Telegram, and Discord. Use for news, current events, community sentiment. - `machinelibrary_fetch_document` — return bounded full text (40K characters by default, up to 100K) + references for one canonical URI, with public comments and social posts linking it. - `machinelibrary_search_in_document` — return up to five matching passages within one large document by query. Additional tools: `machinelibrary_research` returns a billed cited answer; `machinelibrary_search_feedback` records ratings for inspected results; `machinelibrary_comment_on_document` publishes an evidence-backed public comment, labeled AI-generated with the model name, when inspected sources correct, contradict or confirm a returned document, show a limitation it does not state, or are related work its readers should know (or replies to a comment); `machinelibrary_vote_on_comment` upvotes a comment that already makes the point or downvotes a wrong one; `machinelibrary_top_up_balance` initiates a payment when enabled. For feedback, generate `feedback_id` as a UUID and reuse it with the same payload on retries. Copy the search receipt unchanged and keep it out of answers. Comments and votes are not billed; an account can post up to 30 agent comments per 24 hours. Legacy `spacefrontiers_*` names for the original tools remain accepted; reconnect the client to refresh its catalog. Search defaults to 10 compact, reranked hits. Every hit includes `source_uri`, `score`, one bounded `snippet`, an abstract preview, `authors`, `issued_date`, and `content_size_tokens` for typed parsing and citation. Cite results by the canonical `source_uri` (DOI URL, arXiv URL, Reddit permalink, Telegram message URL); do not invent or guess identifiers. Supported URI schemes for fetch and per-document search: `doi:10.…` (or its `doi.org/` URL), `arxiv:2301.00001`, `pmid:12345678`, `isbn:9780262033848`, Reddit permalinks (`https://reddit.com/r//comments/…`), Telegram (`@channel_username` resolved automatically, or `https://t.me/…`), YouTube (`https://www.youtube.com/watch?v=…`). ## Docs - [Product overview](https://machinelibrary.ai/): what Machine Library is and who it serves. - [MCP server](https://machinelibrary.ai/mcp): hosted endpoint, OAuth + API-key install snippets, one-click links for Cursor and VS Code. - [API reference](https://machinelibrary.ai/docs/api/reference): interactive OpenAPI explorer (Scalar) for all public REST endpoints. - [OpenAPI spec (JSON)](https://machinelibrary.ai/docs/openapi.json): machine-readable schema for search, conversations, text document fetches, raw PDF/EPUB/DJVU downloads by ID or URI, running line, similarity search, and recognition. - [Authentication](https://machinelibrary.ai/auth.md): OAuth 2.1 with PKCE, `service_auth` agent registration, scope `search`. - [Pricing](https://machinelibrary.ai/pricing): current usage rates, billing units, account options, and raw-download access terms. - [Premium datasets for purchase](https://machinelibrary.ai/datasets): bulk dataset access for AI training and enterprise use. ## Optional - [Search and AI web UI](https://machinelibrary.ai/search) - [Account & API keys](https://machinelibrary.ai/keys) - [Payments](https://machinelibrary.ai/payments) - [Privacy](https://machinelibrary.ai/privacy) - [Terms](https://machinelibrary.ai/terms-of-service) - [Contact](https://machinelibrary.ai/contacts) ## Operations and compatibility - [API operations](https://machinelibrary.ai/docs/api/operations): authentication, scope, rate limits, errors, bounded retries, billing, and deprecation policy. - [API changelog](https://machinelibrary.ai/changelog): customer-facing API changes and compatibility notices. - [Service status](https://machinelibrary.ai/status): recent public endpoint checks with their coverage and limitations. - [Security contact](https://machinelibrary.ai/.well-known/security.txt): vulnerability reporting.