About Lune Research
Lune connects AI agents with research knowledge and tools built for scientific workflows. Add Lune to an AI app such as Claude, Codex, Cursor, or VS Code, then ask a research question in plain English. The agent can search top-tier conference papers, inspect abstracts and metadata, follow citation trails, read full text when available, and consult curated guidance for the work around the paper.
Why we built it
AI can produce a plausible answer without checking whether the literature supports it. That is not enough for a literature review, a novelty claim, an evaluation plan, or a technical decision with real consequences. Lune retrieves the research first. It gives the agent sources it can parse, compare, and cite, so a reader can inspect where the answer came from.
How Lune grounds an answer
A plain-English question becomes a search across the Lune corpus. Results include paper titles, authors, venue and publication metadata, abstracts, citation relationships, and stable source links. When full text is available, the agent can retrieve the relevant sections and check the passages behind a claim. When the corpus does not contain enough evidence, the answer should say so instead of filling the gap from model memory.
How the corpus works
Lune focuses on top-tier academic conferences selected with established CORE and CCF rankings. It builds the corpus from published conference metadata, open-access repositories, and citation records. Where lawful full text is available, the ingestion pipeline parses and indexes it for retrieval. Lune does not redistribute paywalled articles. Public paper pages expose rights-safe metadata and abstracts, while authenticated tools apply the permissions and usage rules attached to each credential.
Guidance for scientific workflows
Finding papers is only part of research. Lune also maintains a curated guidance library for literature reviews, evaluation design, ablation studies, peer review, research writing, and venue selection. An agent can retrieve this guidance alongside the literature, which helps researchers and technical teams move faster from an early idea to a decision or draft that is ready for scrutiny.
For high-stakes work, Lune keeps sources visible, separates evidence from inference, and gives the agent research-specific tools instead of asking it to improvise a scientific workflow.
How to use Lune
The product is the toolset. Agents connect through the hosted Streamable HTTP MCP server, an official connector or plugin, or the published local MCP package. The Lune command line app can configure local clients. The underlying REST API is not a supported public integration path. The Assistant, Critique, and Workspace pages in the dashboard are working examples of the same tools, not separate sources of evidence.
Who operates Lune
Lune is operated by Retrograde Labs, a research and software company. Product support, privacy requests, and legal notices use separate published contact channels so each request reaches the right owner. See the contact page for those addresses, the Privacy Policy for data practices, and the developer index for connector, plugin, and MCP setup.