Agents · Research methods
How Researchers Use Agents to Manage Literature
The value of an Agent is not collecting more papers. It is working with the context already in the library. Put that collaboration into import, reading, and writing — not into a separate chat product.
Import and organize
Import from identifiers, links, the browser connector, or Zotero. An Agent can extract key points, suggest tags, and draft an initial note. You still decide whether a paper stays, which folder it belongs in, and which attachments are worth keeping.
Read and ask better questions
Highlight, annotate, translate, or ask inside a PDF, then connect passages to papers, concepts, and ideas. Context should not die with the thread. The next time you open the paper, the last question is still there.
Carry results into writing
Markdown notes and BibTeX citations continue into LaTeX, reports, and the next search. You choose what the Agent sees and what becomes part of the knowledge base. The Agent speeds up organization. It does not replace citation or argument.
BYOA: use the Agent you already configured
Connect a local Agent over ACP instead of handing the library to one cloud vendor. Prompts, opted-in files, and credentials stay on your machine. Changing models does not mean changing libraries.
Control matters more than cleverness
Bulk summaries look smart and can write errors into notes. A better default: the Agent proposes, you accept, then the file changes. Automatic close reading is opt-in, not a hidden pipeline.
Frequently asked questions
- Do I need the CLI to use an Agent?
- No. The desktop app calls an installed Agent from the reader. The CLI is for scripts and batch jobs, not a prerequisite.
- Will the Agent rewrite my Vault on its own?
- Not by default. You choose what enters context and which answers become notes.