Skip to content
SciContext

Cloud scientific-paper MCP · Public access in preparation

Give your AI agent papers it can actually cite.

SciContext is preparing a cloud MCP that will let agents search or resolve real papers, query one paper for focused passages, and request a compact provenance receipt when evidence needs an audit trail.

Public MCP coming soonExplore the planned workflow

SciContext is designed to handle paper retrieval inside your agent workflow. Your agent keeps control of reasoning and synthesis.

query workflowIllustrative
search_papers
→ ({ query: "transformer attention" })
← paper.sci_id → scictx:paper:…
// or resolve_paper({ ref: "arxiv:1706.03762" })
query_paper
→ ({ sci_id, query: "How is scaled dot-product attention computed?", prov: true })
← ready → passages + rights notice + one compact receipt
passage benchmark

How is scaled dot-product attention computed?

status
ready
passages
5 returned
rights
CC-BY-4.0
answer term
top passage
Selective provenance · PROV-JSON

Planned public MCP tools

Find or identify. Then ask.

At public launch, search and resolve will be two entry points to a sci_id that agents can question with query_paper.

Two ways in

Start here for discovery

search_papers

Search non-retracted open-access papers by meaning when you do not have an identifier yet.

Start here with an ID

resolve_paper

Resolve a DOI, PMID, PMCID, OpenAlex, arXiv, or SciContext reference to stable paper metadata.

both yieldsci_id
Ask the paper

query_paper

At public launch, agents will ask one focused question and retrieve up to five paper-scoped passages when full text is supported. Add prov: true when your agent needs one compact provenance receipt.

Launch paper providers.

As an independent project, SciContext will launch with arXiv papers in one consistent, citable interface. More providers will follow where they add useful coverage.

Planned for launch

  • arXivPreprints; passages when full text is available

Next paper providers

  • PubMed Central (PMC)Planned
  • Europe PMCPlanned
  • COREPlanned
  • bioRxiv / medRxivPlanned
  • OpenReviewPlanned

Metadata providers.

SciContext combines complementary scholarly metadata to improve paper identity, bibliographic details, topic discovery, citation links, and integrity checks—so agents receive cleaner, better-grounded results.

Planned for launch

  • OpenAlexDiscovery, paper identity & topics
  • CrossrefDOI metadata & integrity signals
  • DataCiteDOI registry metadata
  • OpenCitationsOutgoing reference links

Next metadata provider

  • UnpaywallPlanned