Search full paper text with local embeddings and keyword matching, then rank results with reciprocal rank fusion.
which paper described that method?
Local-first reference manager for macOS
Search the papers you actually read.
paperd keeps your research library on your Mac, turns PDFs into searchable Markdown, and connects Claude to your papers through MCP.
A library built for finding, citing, and revisiting.
paperd combines full-text semantic search, bibliographic metadata, citation context, and AI access without making your papers leave your machine.
Convert PDFs into AI-friendly Markdown while keeping original files and rebuildable metadata as the source of truth.
Let AI clients search, cite, and read your library through a bundled MCP server that works outside the app.
Explore references and citations around each paper to recover context and follow related work.
Resolve papers through arXiv, Crossref, Semantic Scholar, and OpenAlex, then copy clean BibTeX when you write.
Embeddings and PDF parsing run locally. Network access is limited to metadata and paper retrieval you initiate.
From paper to answer.
The workflow is designed around the repeated loop of adding a paper, finding the right passage later, and turning that evidence into citations.
Add
Import by DOI, arXiv ID, URL, or a local PDF.
Index
Convert, chunk, embed, and index the full text locally.
Search
Ask in natural language and jump into the matching section.
Cite
Use BibTeX, notes, citation graph context, or MCP tools.
Install paperd.
paperd is built for macOS 14 or later on Apple Silicon. The local AI models are downloaded during first setup.
brew install --cask paperd-app/paperd/paperd