Agents search and read papers so their hypotheses build on existing work. No account or API key is needed. Turn individual sources on or off from the Data sources menu in the chat composer.
orx discover keyword "speculative decoding" orx discover embedding "faster LLM inference without quality loss" orx discover openalex "protein language models" --published-after 2024-01-01 orx discover pubmed "CRISPR off-target" --prioritize recency --limit 30
Every search accepts --published-after, --published-before, --limit (up to 200), and --prioritize (recency, popular, or historical).
orx paper 2401.12345 orx paper https://arxiv.org/abs/2401.12345 --full orx paper 10.1101/2024.01.01.123456
The source is detected from the ID: arXiv IDs and URLs use alphaXiv, bioRxiv DOIs use bioRxiv, W… IDs use OpenAlex, and PMIDs use PubMed. Override it with --source. alphaXiv returns a paper report; --full returns the full text instead. Other sources return metadata and the abstract.
Create a project From a paper to start from its code. On macOS, the Open in OpenResearch button on a paper's reproduce page at openresearch.sh opens a new project for it in the app.