AI BriefingOpenResearch: Automating Experiments by Using Coding Agents as Research Assistants
alphaXiv/OpenResearch
About the project
Transform coding agents such as Claude Code, Codex, OpenCode, and Cursor into dedicated research workspaces. Help agents autonomously perform the entire process, from literature review and hypothesis formulation to experiment execution and result analysis.
Explore multiple research directions in parallel using independent agent sessions and isolated git worktrees. All experiments are recorded in a git-based tree to ensure reproducibility, with logs and code changes linked to the results.
Run the same code snapshot on various infrastructures, including local environments, SSH, Slurm, Kubernetes, and Ray. Data and experiment records are stored by default in a local SQLite database, maintaining privacy without external exposure.
Connect local models such as LM Studio and Ollama, or custom endpoints, for flexible configuration. Researchers can run a loop to validate ideas proposed by the agent and determine the next steps without writing code themselves.
alphaXiv/OpenResearch
Turn your coding agents into research agents
Rust
This introduction was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report errors, attribution issues, or removal requests via Contact.