OpenResearch, from the team behind alphaXiv, turns your coding agents into research agents. It gives agents such as Claude Code the tooling to do real literature work — searching papers, reading them, and synthesizing findings — instead of answering from memory. The project is open source under the MIT license, has earned about 7,400 stars on GitHub, and is actively developed in 2026.
alphaXiv built its name as a place to discuss and explore research papers, and OpenResearch packages that research capability for the agent era. The idea is that the coding agent already on your machine is a powerful worker but a weak researcher: it cannot browse the literature, compare related work or ground a design decision in what has actually been published. OpenResearch connects it to structured paper search and reading, so tasks like “survey approaches to X,” “find the state of the art for Y,” or “what did this paper actually show?” become agent workflows with sources attached. That grounding matters anywhere technical decisions depend on the literature — ML engineering, research prototypes, thesis work and technical due diligence — and it moves the agent’s answers from plausible-sounding recollection toward evidence you can check.
OpenResearch is free and open source under the MIT license.
OpenResearch is for researchers, ML engineers and students who want their coding agent to do grounded literature research — searching, reading and citing real papers as part of its work.