K SC Miao and colleagues introduce Paper2Agent, a multi-agent pipeline that converts a research paper and its associated codebase into a Model Context Protocol (MCP) server and, thereby, into an interactive 'paper agent' that can answer questions, reproduce analyses, apply methods to new data, and interoperate with other paper agents (Miao et al., 2026). The contribution is not merely another scientist agent. It is a proposed change in the unit of scientific communication: from a passive document to an agent-native knowledge object. This review performs an anatomical reading of that claim. It reconstructs the six-step conversion pipeline; inventories the empirical results across AlphaGenome, Scanpy, TISSUE, a 100-paper computational-biology corpus, non-biology computational papers, and a multi-agent causal-gene case at a psoriasis locus; and situates the work against twelve neighbouring architectures of executable and agentic science. Strengths include locked, tutorial-validated tools, a clean separation of tools, resources and prompts, and a credible reduction in code hallucination relative to repository-browsing baselines. Weaknesses include a 26 per cent agentification failure rate on a naturalistic computational-biology sample, residual dependence on Claude Code scaffolding, a validation regime that measures faithful execution more securely than analytical validity, and an under-specified political economy of maintenance, attribution and security. Also available on Academia.edu: https://www.academia.edu/attachments/134344598/download_file?s=portfolio