Web Researcher MCP

Key Features

• •verify_recommendation: Automatically detects when a source ranks itself first in a list, checks for conflict of interest via author bios, and evaluates domain reputation. This ensures that AI recommendations are not secretly self-promotional, providing transparent flags like self_promotion and conflict_of_interest.
• •verify_citation: Verifies whether a citation exists by querying Crossref and OpenAlex, checks retraction status via Retraction Watch, and confirms link liveness with Wayback fallback. When a claim is provided, it performs lexical coverage analysis to detect mischaracterized sources, preventing AI from citing sources that do not support the given assertion.
• •audit_bibliography: Runs verification checks across an entire reference list of up to 200 entries in a single call. Accepts CSL-JSON, RIS, or BibTeX formats and returns a summary with counts for retracted, deadLink, notFound, unchecked, mischaracterized, and ok entries, enabling rapid bibliography audits before publication.
• •Lens search system: Restricts web searches to curated lists of trusted domains such as journalism, academic, government, and clinical. Uses site: operators to exclude non-authoritative sources, so GEO-optimized brand blogs and sloptimized content cannot appear in results. Sixteen lenses are available out of the box.
• •Research sessions: Creates a named context that records every URL visited, search run, and source scraped during a research task. The session can be exported as a full bibliography in multiple formats and then audited for trust, providing a verifiable audit trail that cannot be manipulated.
• •Domain reputation classification: Every source scraped is labeled with sourceType, authorityTier, and domainReputation from a curated dataset. A peer-reviewed journal gets authorityTier: high, while a company blog gets authorityTier: low. These labels travel with search results so AI assistants can see the trustworthiness of each source.

Use Cases

• •A business executive evaluating project management software uses Web Researcher MCP to verify the chatbot's top recommendations. verify_recommendation reveals the first source is self-promotional from the vendor's blog, and verify_citation finds the surviving source does not support the enterprise claim. Independent journalism lens searches provide genuine reviews.
• •An academic researcher submits a paper with 150 references. Before submission, they run audit_bibliography on the BibTeX file and discover two retracted papers and three dead links. The tool's unchecked category alerts them to a paywalled report that could not be verified, prompting them to replace it with a confirmed source.
• •A journalist investigating a company's claims uses verify_citation on three cited sources. One source returns a 404 but the Wayback fallback provides an archived version showing the page never contained the claimed statistic. The mischaracterized flag prevents publishing a false narrative.
• •A lawyer preparing a brief uses verify_citation on all cited legal cases. The tool flags one DOI as not_found after checking Crossref, indicating a possible fabricated citation. This prevents a costly sanction for filing with fake references, as documented in Stanford's law review.
• •A content writer researching a competitor comparison uses web_search with lens: 'tech' and 'journalism' to find independent reviews. The results exclude the company's own marketing blog due to the domain reputation filter, ensuring the comparison is based on genuine third-party evaluations.

Who is this for?

Developers integrating AI assistants via MCP who need reliable source verification. Academic researchers and graduate students verifying citations in papers and theses. Journalists fact-checking sources for articles. Lawyers and legal professionals auditing citations in briefs and filings. Executives and procurement teams evaluating vendor claims. Data scientists curating training datasets. Editors and publishers ensuring citation integrity in published works.