agent-papers-cli

agent-papers-cli addresses the challenge faced by AI agents (especially those in native bash environments like Claude Code) in autonomously discovering, acquiring, and reading academic literature during in-depth research. It integrates two core CLI tools—paper and paper-search—to bring paper search, PDF reading, content extraction, and metadata retrieval into the command line. This enables agents to independently complete the full research pipeline, from searching for literature and downloading PDFs to reading abstracts, studying specific sections, and tracking citations, much like a human researcher.

Usage is straightforward. Developers or agents simply run simple commands in the terminal, such as paper-search "query" to search for papers, paper read <arxiv_id> to read the full text or specific sections, or paper outline to quickly browse a paper's structure. The tool supports various citation formats including arXiv ID, DOI, and PDF file paths, and can fetch metadata and BibTeX citations from multiple academic databases (e.g., Google Scholar, Semantic Scholar).

This toolkit is ideal for users who need AI agents to perform automated, in-depth literature reviews. For example, academic researchers can use it to quickly build a material library for literature surveys; engineers developing agent applications can use it to equip agents with the ability to "consult literature"; and data analysts or industry researchers can use it to efficiently track the latest developments in specific fields.

Regarding usage recommendations, for users aiming to implement a fully automated research workflow, it is advised to combine the tool with the four pre-built Claude Code skills provided in the repository (such as deep investigation and systematic literature review), which demonstrate how to combine the basic tools into complex multi-step workflows. Note that some advanced features (e.g., chart detection) require installing the additional [layout] dependency package.