Overview

Cito is a hybrid search engine built over the Semantic Scholar corpus, containing over 236 million academic papers. It fuses keyword indexing with dense vector search and cross-encoder reranking to provide precise literature discovery. Designed for AI agents, it features a plain JSON API and a native MCP endpoint that allows tools like Claude Code to perform deep research without facing restrictive upstream rate limits.

Founder story

Cito was launched in 2026 as a solution to the frustration of existing academic APIs throttling AI agents during deep research tasks. Built by researchers for researchers, it provides a performant, agent-friendly alternative that keeps the Semantic Scholar corpus accessible and searchable for automated workflows.

What it does

• • Provides semantic search over 236 million papers using SPECTER2 dense vectors and cross-encoder reranking
• • Offers a native MCP endpoint enabling AI coding agents to execute deep literature reviews autonomously
• • Features a plain JSON API for programmatic access to academic data without complex authentication overhead
• • Allows free web-based literature exploration with no sign-up requirements for individual researchers
• • Delivers high-precision research results by combining traditional keyword indexing with modern neural search techniques.

Who it's for

• AI Developers
• Academic Researchers
• Data Scientists
• Literature Reviewers

Why it works

• Native MCP support allows AI agents to bypass traditional API rate limits for deep literature research
• Hybrid search architecture ensures both high recall and relevance by fusing keyword and dense vector indices
• JSON-first API design simplifies the integration process for developers building agentic research workflows
• Unrestricted access model for the web interface promotes rapid adoption and ease of use for students and academics
• Vector search optimization enables the entire corpus of 236 million papers to be served efficiently on a single infrastructure box.

Growth strategies

• Targeting the growing agentic AI ecosystem by providing a rate- limit-free research backend.
• Leveraging the open- access academic research community to drive organic usage and visibility.
• Integrating directly into popular AI development toolsets like Claude Code and Cursor.
• Differentiating via technical performance (cross- encoder reranking) versus generic LLM-based search engines.

Alternatives

• Consensus
• Scispace
• exa.ai
• Logically

Comparison overview

• Cito is optimized for agentic, rate-limit-free research via MCP, while competitors primarily focus on human-facing chat interfaces
• [Cito provides a JSON/MCP API for deep agent research , while alternatives often restrict programmatic access or prioritize conversational synthesis over raw literature retrieval/]