Loading ToolBench
MCP server for DAO-centred research: Zenon DAI, IAA Publications, ADAJ, plus a curated cross-platform paper-search surface (Crossref, OpenAlex, Semantic Scholar, arXiv, CORE, Zenodo, bioRxiv) — every hit pre-renders an inline_citation block for verbatim agent copy.
Server has 6 tools with complete descriptions and parameter schemas. Naming follows verb_noun pattern (search_*). Descriptions are detailed (150-400 chars) and include domain context (e.g., ADAJ year filtering applied client-side). However, parameter descriptions lack explicit format/range constraints (e.g., max_results 1-50 stated but not in schema), and output schemas are not documented in the tool definitions. Error handling is minimal, no recovery guidance or actionable error messages visible. Tool composition is sound (each does one search task), but responses lack pagination/limit enforcement details.
Health check. Returns "pong" if the server is alive. Use this to verify the server is loaded and responsive before issuing real search calls.
Search the DoA Publication Archive — ADAJ (Annual of the Department of Antiquities of Jordan), SHAJ (Studies in the History and Archaeology of Jordan), Munjazat, JERD, Athar, and other DoA series. English-language, primary source for Jordanian excavation reports. Note: ``year_from``/``year_to`` are applied client-side after retrieval (the upstream GET search ignores year parameters). Narrow your query keywords if year filtering yields too few hits. Citation rendering (Schema v2): each returned ``DAOPaper`` carries an ``inline_citation`` block. Copy ``inline_citation.markdown`` verbatim for in-text citations — do not reformat. For the bibliography / reference-list section copy ``inline_citation.authoritative_bibliography_line`` verbatim; if it is ``None`` (venue metadata incomplete) fall back to Author-Year + URL/DOI rather than reconstructing the reference line from training knowledge.
Search arXiv preprints — the canonical preprint server for physics, math, CS, and Digital Humanities methods (RAG, NLP, GIS, 3D reconstruction). Strongest for papers touching archaeology that appear here within days of upload but rarely show up in Zenon or Crossref.
Search bioRxiv (+ optionally medRxiv) preprints via Europe PMC. Strongest for ancient-DNA / paleogenomic / bioarchaeology preprints that sit on bioRxiv 6–12 months before journal publication. The Levant aDNA literature (Lazaridis, Feldman, Harney, Reich Lab) is the canonical use case.
Output schemas not documented. Tool descriptions state what is returned (e.g., 'DAOPaper' with 'inline_citation' block) but JSON Schema output type is not visible in tool definitions. LLMs cannot plan downstream operations without knowing response structure.
Parameter constraints stated in descriptions but not enforced in schema. E.g., 'max_results: 1-50' is prose in description, not a JSON Schema minValue/maxValue. LLMs cannot parse prose constraints reliably.
No error handling or recovery guidance visible. Tools do not document what happens on network failure, empty results, or invalid queries. No actionable error messages or retry logic documented.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 57 | <=2025-11-25 | v2 |
Search CORE — aggregates open-access full-text content from ~10k institutional repositories. Strongest free OA-PDF discovery surface for grey literature and dissertations.
Search Crossref — the canonical DOI registry, ~150M scholarly works. Use for verifying mainstream peer-reviewed Anglophone literature (Cambridge, Wiley, Elsevier, Tandfonline, etc.) by DOI or bibliographic query.
Pagination not documented. Tools accept max_results but do not document offset/cursor, total count, or how to fetch next page. Large result sets risk context window exhaustion.
Tool descriptions include implementation details (e.g., 'year_from/year_to applied client-side') that are not actionable for LLM selection. Should focus on WHEN to use each tool vs similar ones.