OpenRaiser/NanoResearch

Executes end-to-end autonomous research pipelines from idea to LaTeX-formatted paper, with a 9-stage workflow that generates code, submits GPU training jobs to SLURM or local execution, and produces publication-ready figures from real experimental results rather than LLM hallucinations. Integrates with OpenAlex and Semantic Scholar for literature discovery, supports multi-model routing per pipeline stage, and provides both CLI with TUI interface and Claude Code mode for interactive workflows. Outputs complete reproducible research workspaces including training logs, intermediate artifacts, and structured evidence for auditable scientific workflows.

144 stars.