Research Solutions' Scite MCP Gives AI Agents Access to 250M Open Access Papers
Research Solutions Launches Scite MCP, Giving AI Agents Access to 250 Million Scientific Papers
Research Solutions (NASDAQ: RSSS) launched Scite MCP on February 26, an integration that lets AI agents powered by the Model Context Protocol query a database of more than 250 million open access articles, book chapters, preprints, and datasets. The product, built by Scite, a Brooklyn-based AI research discovery platform that Research Solutions acquired in 2023, connects to ChatGPT, Claude, Microsoft Copilot, Cursor, Claude Code, and any other MCP-enabled application, according to the company's investor announcement.
What makes the integration technically distinct is Scite Smart Citations. Unlike a conventional literature search that returns a list of papers, Smart Citations classify each citation's relationship to the claim it references: whether a paper supports, merely mentions, or actually contradicts the finding it cites. For an AI agent conducting a literature review, that distinction is meaningful. It turns a keyword search into something closer to argument tracing.
"Researchers spend enormous time determining whether a cited paper actually supports the claim made in the text," Josh Nicholson, Chief Strategy Officer at Research Solutions and co-founder of Scite, said in the launch announcement. "Scite Smart Citations provide context that has never been available at this scale."
The catch is that Scite MCP currently accesses open access content only. Open access advocates have estimated that around 30 to 40 percent of articles are published under OA licenses, with variation by field. That means the majority of peer-reviewed research sits behind publisher barriers. Elsevier, Springer Nature, Wiley, and other commercial publishers control most of that paywalled corpus. Research Solutions says discussions with publishers are underway, but no deals have been announced.
That limitation defines the product's current utility. A researcher agent working primarily with preprints and OA journals gets a genuinely useful tool. An agent synthesizing the clinical literature on a drug interaction, much of it in paywalled journals, hits a wall.
The MCP framing is worth noting. Rather than building a proprietary agent-facing interface, Scite published a Model Context Protocol server. Any agent that speaks MCP can use it. This is infrastructure positioning: Scite wants to be the literature layer that other agents depend on, rather than a destination tool researchers visit directly. Whether that strategy holds depends partly on whether the publisher deal comes through. Without paywalled access, the infrastructure is incomplete by design.
What Scite MCP represents is a concrete step in the ongoing construction of agent-accessible external data layers. Scientific literature has been a conspicuous gap. Publishers have been slow to offer machine-readable licensing at scale, and even when they do, the technical interfaces are inconsistent. An MCP standard for literature access, if it gains adoption, could change the economics of AI-assisted research and could create new tensions with publishers who have not licensed their content for agentic use.
The publisher question is the one to watch. If Research Solutions closes deals with the major commercial publishers, Scite MCP becomes a significant piece of research infrastructure. If it remains open-access only, it is a well-built tool for a subset of the scientific literature, useful but not the lit layer that agents genuinely need.
Research Solutions acquired Scite in 2023, according to a deal summary on PrivSource. Josh Nicholson is Chief Strategy Officer at Research Solutions and co-founder of Scite.