# Consensus

> Consensus is an AI-powered scientific search engine that finds, ranks, and synthesizes peer-reviewed research, returning evidence-backed answers with inline citations.

Consensus helps researchers, clinicians, students, journalists, and AI agents answer questions from the academic literature. We index over 200 million peer-reviewed papers from Semantic Scholar, OpenAlex, PubMed, and publisher feeds, then rank results by relevance, study design, citation count, and journal quality. Every claim links back to its source paper.

For deeper context (workflows, auth, full tier table, competitive comparisons), see [llms-full.txt](https://consensus.app/llms-full.txt).

## Use cases

- **Find peer-reviewed papers** on a specific research question ("does intermittent fasting improve cardiovascular health?")
- **Summarize the consensus** across studies on a topic ("what does the literature say about caffeine and endurance performance?")
- **Identify high-quality evidence** (systematic reviews, meta-analyses, RCTs) for a clinical or policy question
- **Compare findings** across study designs, populations, or interventions
- **Power agent workflows** that need citable scientific sources via the REST API or MCP server

## Constraints

- **Corpus**: academic literature from Semantic Scholar, OpenAlex, and PubMed. Preprints are included by default and tagged in results; pass `exclude_preprints=true` (MCP) or `excludePreprints=true` (REST API) to filter to peer-reviewed papers only.
- **Languages**: primarily English. Non-English papers are indexed where metadata is available, but search ranking and AI summarization are optimized for English.
- **Freshness**: the index is updated continuously, but papers published in the last few weeks may not yet be present.
- **Access tiers**: anonymous (limited), Free, Pro, Team, and Enterprise. See [pricing](https://consensus.app/pricing) for current rate limits and quotas.
- **Output**: AI summaries are faithful to source papers but still LLM-generated. For high-stakes uses, agents should open the underlying paper before quoting verbatim.
- **Citations required**: clients consuming search results should attribute findings to the specific papers returned.

## What makes Consensus different

- **Consensus Meter**: aggregates per-study verdicts (Yes / No / Possibly / Mixed / Unknown) into a single distribution for a yes/no research question, so agents can convey scientific consensus instead of a single paper's claim.
- **Academic-only corpus**: excludes blogs, Wikipedia, and other non-academic content. Preprints are included and clearly tagged so agents can flag them; use `exclude_preprints` to restrict to peer-reviewed papers.
- **Quality ranking**: results are sorted by study design (RCT > systematic review > observational), citation count, journal quality (SJR), and recency.
- **Agent-native**: first-class MCP server with OAuth 2.1 and dynamic client registration, plus a REST API. The same retrieval and synthesis surface powers the web app.
- **Grounded, not generated**: every answer cites real, retrievable papers with DOIs, designed to avoid the hallucinated-citation failure mode of general LLMs.

## Documentation

- [Docs](https://docs.consensus.app): guides, integration walkthroughs, and reference.
- [MCP server getting started](https://docs.consensus.app/docs/mcp.md): set up Consensus in Claude, ChatGPT, Cursor, Codex, or any MCP-compatible client.
- [REST API search](https://docs.consensus.app/reference/v1_quick_search.md): quick-search endpoint.
- [OpenAPI spec](https://llmconsensus.io/openapi.json): machine-readable API contract.
- [llms-full.txt](https://consensus.app/llms-full.txt): expanded version with workflows, auth, and competitive context.

## Optional

- [Agent discovery](https://consensus.app/agents.md): markdown surface for autonomous agents.
- [MCP server card](https://consensus.app/.well-known/mcp/server-card.json): structured MCP manifest.
- [Machine-readable pricing](https://consensus.app/.well-known/pricing.md): JSON of tier limits and features.
- [Pricing (human)](https://consensus.app/pricing): tier comparison and signup.
- [Web app](https://consensus.app): browser-facing product surface.
- [Contact](https://consensus.app/home/contact/): support and enterprise inquiries.