Research Chain MCP Server

@pipeworx/research-chain

Everything openly available around one academic paper, in a single call: full text, license, datasets, code, and citations. Given a DOI, PMID, or arXiv id, research_chain walks the join that Crossref, OpenAlex, Unpaywall, DataCite, Zenodo, Hugging Face Papers, PubMed Central, and Semantic Scholar each expose piecemeal — so a caller no longer has to know all six/seven tools to answer "what datasets and code are linked to this paper?" or "find an open copy of this DOI and its license."

Part of Pipeworx — an MCP gateway connecting AI agents to 1856+ live data sources. This is an independent, unofficial integration — not affiliated with, endorsed by, or published by the upstream provider.

Tools

• research_chain({doi | pmid | arxiv_id, max_datasets?, max_citing?}) — returns:
  • paper — title, journal, publisher, authors, Crossref's stated license(s) (URL + which manuscript version each applies to), OpenAlex concepts and citation count.
  • open_copies[] — every open-access location Unpaywall knows about, each with its OWN host_type, version (publishedVersion | acceptedVersion | submittedVersion), and license — never collapsed. is_oa: true only means "a free-to-read copy exists somewhere"; it says nothing about reuse rights, which is why license is null (not guessed) whenever Unpaywall doesn't state one. A PubMed Central copy (if one exists) is added as its own entry with license and version explicitly null, since NCBI's ID Converter confirms the copy exists but doesn't state its terms.
  • pmc_full_text — {has_full_text, pmcid, pmid, url} from NCBI's ID Converter.
  • datasets — {total, returned, results[]} from DataCite's relatedIdentifiers search (datasets/software that cite the paper's DOI). total is the real upstream count even when results is truncated to max_datasets.
  • code — {zenodo[], paperswithcode}. Zenodo is searched by metadata.related_identifiers.identifier. paperswithcode (Hugging Face Papers, the de-facto successor) reports {found: false, results: []} when the upstream search comes back empty — an empty result is not_found, never treated as an answer (a bare [] from this upstream used to be routed to as if it were a real "no code exists" answer).
  • citing — {count, top[], note} from Semantic Scholar: count is the paper's total citation count, top is up to max_citing citing papers sorted by their own citation count. note carries an honest partial- failure message (e.g. a rate limit) when one of the two Semantic Scholar calls failed but the other succeeded — it is never silently dropped.
  • retraction — derived from the SAME Crossref record already fetched for paper (title marker, update-to retraction notices, is-retracted-by relation) — no extra API call. retracted: false means "no retraction on record", not proof the paper is clean (coverage depends on the publisher).
  • sources_used / sources_failed — every leg fetches independently and in parallel; a slow or dead upstream lands in sources_failed with a reason, and every other leg still returns.
  • as_of — ISO timestamp of the call.

Auth

Keyless. Every upstream here is a public, unauthenticated API. Crossref, OpenAlex, and the NCBI ID Converter ask only for a polite-pool contact email, which this pack supplies as its own (hello@pipeworx.io) — the same pattern as the crossref, openalex, and unpaywall packs.

Data sources

• https://api.crossref.org/works/{doi} — title, journal, license, retraction markers.
• https://api.openalex.org/works/doi:{doi} — concepts, citation count.
• https://api.unpaywall.org/v2/{doi} — open-access locations, each with version + license.
• https://pmc.ncbi.nlm.nih.gov/tools/idconv/api/v1/articles/ — DOI/PMID ↔ PMCID resolution. NCBI moved this off www.ncbi.nlm.nih.gov/pmc/utils/idconv in 2026 — the old host 301-redirects here; this pack hits the new host directly.
• https://api.datacite.org/dois — datasets/software citing the paper (relatedIdentifiers.relatedIdentifier search).
• https://zenodo.org/api/records — records related to the paper. The search field is metadata.related_identifiers.identifier, not the bare related_identifiers.identifier — the bare field silently matches zero records for every query (verified live: 0 vs 19 hits for the same DOI).
• https://huggingface.co/api/papers/search — code/implementation search (successor to the shut-down Papers with Code API).
• https://api.semanticscholar.org/graph/v1/paper/DOI:{doi} and /citations — citation count + citing papers. Keyless pool is ~1 req/sec cumulative across BOTH endpoints; this pack sequences them with a ~1.1s stagger inside one leg rather than firing both in the same instant, which 429'd one of them on nearly every call before the fix.

OSF was scoped as a third code/materials source but dropped: its public API has no DOI-filterable field (/v2/preprints/?filter[doi]=... returns HTTP 400 "not a filterable field", verified live). Zenodo + Hugging Face Papers cover the code/materials join for now.

Identifier resolution

• doi — used directly (a doi.org URL or doi: prefix is stripped).
• pmid — resolved to a DOI via NCBI's ID Converter.
• arxiv_id — resolved to arXiv's own auto-registered DataCite DOI (10.48550/arXiv.<id>), which covers essentially every arXiv submission since 2022. This is a construction, not a verified lookup — a pre-2022 arXiv id without a registered DOI will report not_found on the downstream legs rather than resolving.

Quick Start

Add to your MCP client (Claude Desktop, Cursor, Windsurf, etc.):

{
  "mcpServers": {
    "research-chain": {
      "url": "https://gateway.pipeworx.io/research-chain/mcp"
    }
  }
}

What this endpoint actually serves

tools/list at https://gateway.pipeworx.io/research-chain/mcp returns the tools in the table above plus the shared Pipeworx meta-tools — ask_pipeworx, discover_tools, search_within, remember/recall and the rest of the gateway-wide set. So the tool count you see is larger than this table: a single-pack endpoint currently lists roughly 30 shared tools alongside the pack's own. The connection's initialize response states its exact scope, and is the authoritative answer for a given day.

This is deliberate, not multiplexing by accident. The meta-tools are what let a scoped connection answer a question this pack does not cover — via ask_pipeworx, which routes across the whole catalog — without you adding a second MCP server. There is currently no way to mount a pack endpoint without them; if the extra schemas cost you more context than the routing is worth, connect to the full gateway once rather than to several pack endpoints.

Or connect to the full Pipeworx gateway to get every pack's tools listed directly, instead of just this one's:

{
  "mcpServers": {
    "pipeworx": {
      "url": "https://gateway.pipeworx.io/mcp"
    }
  }
}

Both URLs reach the same gateway and the same 1856+ data sources. The only difference is which pack's tools are listed directly; ask_pipeworx reaches all of them from either one.

No MCP client? Call it over HTTP

curl -X POST https://gateway.pipeworx.io/v1/tools/research_chain \
  -H 'Content-Type: application/json' \
  -d '{"doi":"10.1038/s41586-020-2649-2"}'

No account needed for the first calls. Inspect any tool: GET https://gateway.pipeworx.io/v1/tools/research_chain. Find one: POST https://gateway.pipeworx.io/v1/tools/search_packs with {"query":"..."}.

Run it from npm

@pipeworx/mcp-research-chain is a local stdio bridge to the Pipeworx gateway: your MCP client starts it with npx, and it forwards every call to the same https://gateway.pipeworx.io/research-chain/mcp endpoint shown above. It needs network access; there is no offline mode.

{
  "mcpServers": {
    "research-chain": {
      "command": "npx",
      "args": ["-y", "@pipeworx/mcp-research-chain"]
    }
  }
}

Without a key it calls anonymously, capped at 50 calls a day. A free account raises that to 200 a day: get a key at https://pipeworx.io/signup and add "env": { "PIPEWORX_API_KEY": "<your key>" } to the server entry. The key is sent as a Bearer token, the same way a hosted connection sends it, so calls are metered and billed exactly as they are there.

It lists only this pack's tools by default. Set PIPEWORX_META_TOOLS=1 to also list ask_pipeworx and the other gateway-wide tools.

Behind an HTTPS proxy: Node's built-in fetch does not read HTTPS_PROXY by default. Newer Node releases honour it when NODE_USE_ENV_PROXY=1 is set; check your Node version's documentation.

Using with ask_pipeworx

Instead of calling tools directly, you can ask questions in plain English — this works on the pack endpoint above as well as on the full gateway:

ask_pipeworx({ question: "your question about Research Chain data" })

The gateway picks the right tool and fills the arguments automatically.

More

• Docs and guides
• pipeworx.io

License

MIT