Retrieve a web page or file from the user's own machine, reproducing the user's access (their IP/VPN and a real browser User-Agent) so resources that block generic crawlers still work. Runs a retrieval ladder (curl → wget → headless Chrome) and returns HTML as clean Markdown, JSON/text as-is, and saves PDFs/binaries to disk with a reported path. Use this to read sources that a direct client cannot reach; pair it with web search, which finds the URLs to fetch.

Search the open web for sources scholarly indexes miss: grey literature, official/regulatory documents, press releases, analyst commentary, and news. Essential for novel or fast-moving topics (e.g. an emerging AI legal matter) where peer-reviewed literature is thin. Keyless (DuckDuckGo) and retrieved server-side. Returns title, URL, and snippet per result — then retrieve with fetch-url and weigh with appraise-source (mind the source-quality cadence). For peer-reviewed work, prefer search-literature.

Search the scholarly literature across one or more free, keyless indexes: OpenAlex (default, ~250M works with citations), Crossref (grey literature, very current), arXiv (CS/AI/physics/quant preprints), and OSF (SocArXiv/PsyArXiv preprints). Results are merged and de-duplicated by DOI/title and ranked so cited, peer-reviewed work sits above zero-citation preprints (matching the source-quality cadence). Each result is tagged with its source and shows title, authors, year, venue, citations, OA status, DOI, and ID. Use the sources array to widen reach; pass results to appraise-evidence, fetch-paper, or export-bibliography. For open-web/grey sources beyond academia, use web-search.

Resolve a paper by DOI, OpenAlex ID, URL, or title and return its metadata plus, where open access, its full text. Looks up OpenAlex for metadata and the open-access URL, retrieves the full text via the fetch-url ladder, and parses PDFs in-process. For paywalled papers it returns metadata and the best available link. Feed the result to appraise-evidence or export-bibliography.

Extract text from a PDF for appraisal or data extraction. Accepts a local file path or an http(s) URL (retrieved server-side via the fetch-url ladder, so paywalled-but-entitled or bot-protected PDFs work). Returns the extracted text in-process — no external converters required. Pair with appraise-evidence to critically appraise a retrieved paper.

Map the citation neighbourhood of a seed paper using OpenAlex: its backward citations (the works it references) and/or its forward citations (the works that cite it), each ranked by citation count. Use to snowball a literature search, trace a finding to its origins, or surface the most influential follow-on work. Resolve the seed by DOI, OpenAlex ID, URL, or title.

Look up a researcher on OpenAlex and return their profile: works count, total citations, h-index, last-known affiliation, ORCID, and their most-cited works. Use to gauge a source author's expertise and track record, or to find a researcher's key papers.

Build a bibliography from a list of papers (DOIs, OpenAlex IDs, URLs, or titles) in BibTeX, RIS, or CSL-JSON. Uses DOI content negotiation for publisher-grade entries and falls back to OpenAlex metadata when a work has no DOI. Output imports cleanly into Zotero, EndNote, or Mendeley, or drops into a LaTeX/Pandoc pipeline. Pair with search-literature to assemble a reference set.

Search the user's local Zotero library via the Zotero desktop app's local API. Answers "what do I already have on X?" before searching the wider literature. Returns matching items with title, authors, year, type, DOI, and Zotero key. Requires Zotero to be running with the local API enabled.

Resolve a paper by DOI, OpenAlex ID, URL, or title and save it to the user's local Zotero library as a journal-article item (title, authors, year, venue, DOI). Requires Zotero running with the local API enabled (Zotero 7+ for writes). Pair with search-literature or fetch-paper to capture a source you want to keep.

Search the configured Obsidian vault (Markdown notes) for a term and return matching files with line numbers and snippets. Use to find existing notes, lit reviews, or evidence tables before creating new ones. Requires RESEARCH_MCP_OBSIDIAN_VAULT to be set.

Read a Markdown note from the configured Obsidian vault by its vault-relative path. Requires RESEARCH_MCP_OBSIDIAN_VAULT.

Write a Markdown note into the configured Obsidian vault — e.g. a literature note, evidence table, or research summary. Modes: create (fail if exists), overwrite, or append. Paths are confined to the vault. Requires RESEARCH_MCP_OBSIDIAN_VAULT.

Turn a fuzzy idea into one or more sharp, researchable questions. Returns a structured scaffold to complete: a FINER worth-pursuing appraisal, the best-fit question framework (PICO/PICOC/PECO/SPIDER/PCC) broken into elements, surfaced constructs and assumptions, scope boundaries, and the question type that points toward a design. Use at the very start of a research effort, before design-study.

Recommend a fit-for-purpose study design for a research question and its constraints, then surface the validity threats it must address. Returns a scaffold: question-type classification, a shortlist of candidate designs with trade-offs, a four-validities threat checklist (internal, external, construct, statistical conclusion), sampling considerations, and a reasoned recommendation. Use after scope-question.

Draft or critique a measurement instrument — a survey/scale, interview protocol, or observation schedule — with a built-in reliability and validity plan. For quantitative scales it walks the DeVellis development sequence, item-writing rules, response-format choice, reliability/validity strategy, and total-survey-error control; for qualitative work it builds a non-leading interview guide with a trustworthiness plan. Provide existing items to critique instead of starting fresh.

Recommend an analysis strategy and keep the inference honest. For quantitative data it matches the question and variable types to a test/model family, lists the assumptions to check, insists on effect sizes with confidence intervals over bare p-values (per the ASA statement), flags p-hacking/HARKing/multiplicity risks, and handles missing data. For qualitative data it selects an analytic approach (e.g. reflexive thematic analysis) with a rigour plan. Use after design-study, before collecting or analysing data.

Critically appraise a single study and return a structured scorecard. Classifies the design, applies the matching risk-of-bias framework (RoB 2 / ROBINS-I / observational / CASP), assesses the results by magnitude and precision rather than significance alone, rates certainty using GRADE, and judges applicability to a decision context. Feed it an abstract or full text (e.g. retrieved via fetch-url).

Appraise a non-study source (press release, analyst note, legislation, news, blog, official statistic) for credibility and provenance — the companion to appraise-evidence. Places it in the 7-tier source-quality cadence, runs CRAAP and SIFT checks, flags interest/bias, and recommends a weight (High / Moderate / Low / Orientation-only) and a one-line provenance label. Use for grey literature and for novel or fast-moving topics where peer-reviewed evidence is thin.

Structure a synthesis across multiple studies. Recommends a review type and synthesis logic (aggregative vs configurative), lays out an evidence/extraction table, assesses clinical/methodological/statistical heterogeneity before any pooling, selects a synthesis method (meta-analysis, thematic/framework/narrative synthesis), rates the certainty of the whole body of evidence with GRADE, and identifies gaps. Provide the studies, or use it to scaffold a systematic review from the protocol stage.

Validate a manuscript or draft against the right reporting checklist (PRISMA, CONSORT, STROBE, COREQ, TRIPOD, or AGREE II), selected from the study design. Returns an itemised pass/gap report — each required item marked Reported / Partial / Not reported / N/A with its location and the gap to fix — plus a summary of the critical omissions. Use before submission or as part of peer review.

Run a rigorous, constructive peer-review pass over a manuscript. Produces a structured reviewer report: summary of the contribution, significance/novelty, prioritised major and minor issues, a methodological critique across the four validities, reproducibility and ethics checks, reporting-standard compliance, and a clear recommendation (accept / minor / major / reject). Composes appraise-evidence, check-reporting, plan-analysis, and assess-ethics for deep dives.

Run a research-ethics and integrity review. Applies the Belmont principles (respect for persons, beneficence, justice), Declaration of Helsinki obligations, research-integrity pillars, data-governance and privacy checks, open-science commitments, and — where AI is used in the research itself — responsible-AI-in-research disclosure. Returns a structured assessment with required approvals, risk flags, and mitigations. Use as a pre-study ethics pre-check and within peer-review.

Convert research findings into a decision-maker brief tuned to an audience (board, executive, HRLT, line managers, or general). Built on evidence-based management: it frames the science alongside organisational data, stakeholder values, and practitioner judgement, communicates confidence honestly, and lands a clear "so what / now what". Produces a BLUF-led brief scaffold; pair with the internal-comms skill for final house-style formatting.

Frame a people-analytics study with a domain preset (engagement, attrition, selection, DEI, wellbeing, or performance). Returns the construct definition, validated instruments to license, example metrics, design considerations, common pitfalls, data sources, and people-data governance flags — then hands off to design-study, develop-instrument, and assess-ethics. Use to start any HR/workforce research so the methodology core is tuned to people data.

Assemble a workforce survey from validated-style construct blocks (engagement, inclusion, psychological safety, manager support, intent to stay, wellbeing). Returns items, response scales, and the validated instrument each block is modelled on, plus survey-hygiene and people-data governance notes. Items are original illustrative wording — license the named instruments for defensible measurement. Refine with develop-instrument.

Review an AI or people-data use case against defensible governance dimensions aligned with NIST AI RMF, OECD, ISO/IEC 42001, and EU AI Act risk tiers. Classifies the risk tier, works a governance checklist (purpose, lawful basis, transparency, fairness, human oversight, validity, privacy, security, contestability, monitoring, documentation), flags when to escalate to an AI-governance committee, and recommends a path (proceed / proceed with conditions / escalate / do not proceed). Complements an organisation’s committee Terms of Reference rather than replacing them; pair with assess-ethics.