Paper Chaser MCP

What Paper Chaser MCP MCP server does

Paper Chaser MCP MCP server exposes academic research workflows through a FastMCP interface. Its guided entrypoint, research, can route discovery, known-item recovery, citation repair, and regulatory questions. The server also supports direct provider searches, paper metadata lookup, citation and reference expansion, paper-title autocomplete, and saved-session analysis.

The project connects to Semantic Scholar, arXiv, OpenAlex, CORE, Crossref, and Unpaywall for scholarly information. Its supported public-sector sources include ECOS, FederalRegister.gov, and GovInfo. SerpApi Google Scholar and ScholarAPI are available as optional provider paths, with separate credentials or paid access required where applicable.

How it works

The default guided profile presents a smaller surface intended for agents that need a clear research path. It exposes research, follow_up_research, resolve_reference, inspect_source, and get_runtime_status. The expert profile adds provider-specific searches, bulk and cursor-paginated operations, graph expansion, regulatory direct tools, and broader diagnostics, subject to enabled features.

A research call can return a reusable searchSessionId. That identifier anchors follow-up questions, result-set quality checks, thematic clustering, and citation-graph expansion. Responses can include answerability, evidence, leads, routing information, evidence gaps, structured sources, and failure summaries.

resolve_reference accepts inputs such as DOI values, arXiv identifiers, URLs, citation fragments, title fragments, and regulatory references. inspect_source provides provenance and trust information for a selected result, while get_runtime_status reports the active profile, transport, provider state, and warnings.

Setup and configuration

The documented local installation path uses the repository source and an editable Python install:

pip install -e .

An MCP client can start the server in stdio mode with python -m paper_chaser_mcp. The guided profile is the default; the example recommends hiding disabled tools.

A .env.example file can be copied to .env for shell or Docker Compose use. Provider-specific credentials should be supplied only for services being used. SerpApi Google Scholar is opt-in and paid, and the SerpApi and ScholarAPI search tools require their respective services.

Tools and capabilities

Paper Chaser MCP MCP server includes capabilities for:

• Smart, brokered, provider-specific, bulk, and entity-constrained paper searches.
• Citation repair, known-item matching, metadata lookup, and paper-title autocomplete.
• Citation and reference traversal through Semantic Scholar and OpenAlex.
• Grounded questions, comparisons, clustering, and research-landscape mapping over saved results.
• Citation-graph expansion across paper and author anchors.
• Runtime, provenance, trust-state, and source-inspection reporting.

Limitations and notes

Search coverage and result quality depend on the enabled providers and their available metadata. SerpApi and ScholarAPI are not part of the basic provider path and require their own access. Title-only or conflicting references can produce multiple candidates or a request for disambiguation; those results should not be treated as citation-ready until resolved. Follow-up operations should use the correct saved searchSessionId, especially when more than one session is available. The README identifies MIT licensing and indicates that PyPI publication remains gated, while GHCR images and GitHub release assets are primary distribution channels.