Agentero is an open-source Zotero alternative that lets agents read your library

A researcher using Zotero and ChatGPT or Claude side by side now has an open alternative that connects the two. poco-ai put Agentero on GitHub under the MIT licence. It is a desktop research workbench for papers and notes. The GitHub repository has 1,495 stars. It was ranked 22nd on Trendshift's daily list on 6 November 2026. The project is written in TypeScript on Tauri.

The project's line, in its own README, is that Agentero is "not meant to replace Zotero". It is "a new way for humans and agents to handle research context together". Humans read, judge and conclude. Agents handle batch retrieval, metadata, figure and formula extraction, and basic summaries, the README says.

What it does

Agentero is a desktop app for macOS, Windows and Linux. It opens PDFs, shows outlines and highlights, and holds Markdown notes with wiki-style links and LaTeX rendering. The README says it parses math formulas, algorithm pseudocode and tables into structured Markdown and LaTeX. It offers selection, side-by-side and full-text translation.

It can import a full Zotero library in one click, keeping tags, notes and attachments, the README says. It works with the Zotero browser extension to save papers from the web. It exports to BibTeX and BibLaTeX. Papers and notes are kept as plain PDF and Markdown files in a regular folder. No private database, no obfuscated layout.

How the AI side works

Two protocols sit under the AI features. The app can talk to a local agent over the Agent Client Protocol. Claude Code, OpenCode, Codex or a local model can then act on the library. The context stays on the user's own machine.

The app can also run as an MCP server. Through the Model Context Protocol, an external client can query the local library. The clients named in the README are ChatGPT Web and Claude Desktop. They pull papers in as context. A headless CLI covers scripted tasks like batch import, metadata extraction and note writing, the README says.

For whom

The target user is someone who already reads papers for a living. A PhD student. A research scientist in a lab. A review author. A policy analyst. Today their flow is Zotero for papers, Obsidian for notes, and a chat window for the model. A lot of copy and paste moves between them.

The MIT licence means a team can use Agentero in a product it sells. The project is built as a desktop app, not a cloud service. The cloud shape is the other common one in this space. A team that cannot send papers to a cloud service can use Agentero locally.

What is not in the README

The README does not list a quality score against Zotero plus community plugins. It does not say how often the figure and formula extraction is right on a hard page. It does not state how many of its releases have been stable. A researcher who switches should keep a paper set they already know, and compare the extraction on those pages first.

Source

• Primary source: poco-ai/Agentero on GitHub.
• Project website: agentero.app.