Loading ToolBench
A research tools MCP server providing search and academic research capabilities through multiple integrated services
The server provides 14 research-focused tools with mostly present descriptions and schemas. However, there are significant gaps in parameter documentation, inconsistent schema completeness, and missing error handling guidance. Average tool description length is ~100-150 chars (within baseline), but many parameters lack descriptions or have vague ones. Output schemas are not documented. Some tools have well-structured inputs (github-search-repositories with 7 params, clear descriptions), while others are minimal (arxiv-get-paper with only arxivId). No tool annotations (readOnlyHint, destructiveHint) despite all being read-only operations. Error handling is absent, no recovery guidance, no validation messaging, no actionable error responses.
Download and process LaTeX source of an arXiv paper
Get detailed information about a specific arXiv paper
Search arXiv papers using various search methods
Perform deep web research on a given topic
Get markdown content from a URL. Use this tool for URLs containing documentation and code examples where precise information is needed.
Search code across GitHub repositories using GraphQL API. Use simple, focused search terms - avoid combining too many keywords as GitHub uses AND logic. Use filters instead of including language names in the query text.
Missing output schema documentation for all 14 tools. LLMs cannot plan downstream calls or extract relevant fields without knowing response structure.
No tool annotations despite all tools being read-only. Adding readOnlyHint=true makes intent explicit and prevents agent misuse.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 50 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 46 | - | v1 |
Search GitHub repositories using GraphQL API. Use simple, focused search terms - avoid combining too many keywords as GitHub uses AND logic. Use filters instead of including language names in the query text.
Search authors using OpenAlex API
Search concepts using OpenAlex API
Search institutions using OpenAlex API
Search academic works (papers, articles) using OpenAlex API
Search Reddit for discussions
Generate research strategies and methodologies
Search for papers on Google Scholar
Minimal parameter documentation in several tools. arxiv-get-paper, arxiv-get-latex-source, scholar-search, and research-strategy have minimal or generic descriptions that do not explain usage context or constraints.
No error handling guidance. Tools provide no recovery instructions, retryability classification, or actionable error messages. Agents cannot self-correct on failures.
Missing or undocumented limits on result counts. Tools like openalex-search-* and arxiv-search accept maxResults but do not document default caps or enforced limits in the implementation.
Limited parameter constraints. Many optional parameters (e.g., searchType, sortBy, sortOrder in arxiv-search) list allowed values in descriptions but are not enforced as enums in schema. LLMs may hallucinate invalid values.