Reimagining research papers as interactive and reliable AI agents - Nature
Paper2Agent automatically transforms static research papers into interactive AI agents that users can query in natural language, eliminating the need to manually install code, configure environments, or parse technical documentation. The system wraps a paper's methods, code, and data as an MCP (Model Context Protocol) server that connects to LLMs like Claude, letting researchers apply the paper's techniques to new problems without programming expertise.
- Paper2Agent solves a real friction point: even well-documented computational methods require substantial setup work (installing dependencies, understanding APIs, configuring parameters), which blocks adoption by researchers without strong technical skills. The system lets users ask questions like "interpret this variant's effect on chromatin accessibility" instead of wrestling with repository setup.
- The framework validates reproducibility by testing each tool against the original paper's reported results and figures, then locks those tools to prevent LLM hallucination and ensure consistent outputs. Every tool includes a code reference back to the original paper for transparency.
- Demonstrated agents successfully reproduced results from AlphaGenome (genomic variant interpretation), Scanpy (single-cell analysis), and TISSUE (spatial transcriptomics), then performed novel analyses like collaborating across multiple agents to prioritize a causal gene for psoriasis.
Paper2Agent is a system that turns published research papers into interactive AI agents that can answer questions and run analyses through natural language. Instead of readers having to hunt down code repositories, install dependencies, and figure out APIs—which creates real barriers for non-programmers—the framework wraps a paper's methods, code, datasets, and workflows into an agent that acts like a virtual corresponding author. The system works by analyzing the paper and its codebase, then building what's called an MCP (model context protocol) server that exposes all the paper's tools in a standardized way that large language models can actually use. A user could ask something like "interpret the effect of this variant on chromatin accessibility" instead of wrestling with AlphaGenome's API hierarchy and parameter setup.
The framework addresses a real problem in scientific communication. Papers are fundamentally passive—you read them, try to understand what the authors did, then attempt to replicate or adapt their work yourself. When methods are computational, this gets messy fast. AlphaGenome is powerful for genome modeling but requires serious technical expertise to deploy. Paper2Agent flips this by making the paper itself executable and conversational. The system uses multiple agents working together to build the MCP server, then tests everything against the original paper's results to catch hallucinations or errors before they happen. Each tool includes a reference back to the original code for transparency.
The researchers demonstrated this with real examples: agents built from AlphaGenome, Scanpy, and TISSUE (a single-cell analysis tool) successfully reproduced the original papers' results and handled novel queries users threw at them. They even showed multiple agents collaborating to identify a causal gene for psoriasis. This isn't just about convenience—it fundamentally changes how research gets shared and built upon. Instead of papers being static documents that sit behind technical barriers, they become interactive collaborators that other researchers can query, adapt, and combine with other paper agents to tackle new problems.