
AutoSearch runs full PRISMA 2020 systematic reviews as MCP tools. Instead of dumping search results and making your agent chase DOIs, it exposes a complete pipeline: question to multi-source search (OpenAlex, PubMed, Semantic Scholar, arXiv, and four others), through screening with recorded decisions, to a finished IMRAD paper with APA references and PRISMA flow counts. Every DOI is verified live against Crossref before it appears in output, and the team publishes a rerunnable citation integrity benchmark. Runs are async, so you launch a deep review, poll for status, and fetch markdown, JSON, or PDF when ready. Reach for this when you need agent-driven literature reviews that produce auditable citations, not just links.