Find your next open-source project

Search by name, purpose, or keyword — or paste a GitHub repository URL

poco-ai/Agentero

Productivity & Utilities

A local-first research desktop app that keeps PDFs, notes and agent context together with citations and MCP access.

1,952
124
TypeScript
MIT
4 months old

What it does

Agentero targets the gap between reference managers and the AI tools researchers now use. Papers sit in one app, notes in another, and model conversations in a third, so users copy-paste context by hand. It also argues that raw PDFs are a poor model input compared with structured Markdown and LaTeX.

The desktop app combines a PDF reader with outline navigation, highlight annotations and full-text search, plus extraction of formulas, algorithm pseudocode and tables into structured Markdown/LaTeX. A built-in Markdown editor supports [[wikilinks]], slash commands and live LaTeX, and notes can be linked to documents. Translation is available as word, page-by-page or full document, with built-in channels or a user-supplied API key.

Agent integration

Agentero connects local agents over the Agent Client Protocol, so Claude Code, OpenCode, Codex or local models can be used without locking into one vendor, and context stays local. It also runs as an MCP server, letting clients such as ChatGPT Web or Claude Desktop query the local library. A headless CLI covers batch import, metadata extraction and note writing for scripts.

For discovery, the app browses Cool Papers and ModelScope with one-click archiving, supports academic RSS, recommends daily arXiv papers from library content, and parses reference lists for batch import.

Getting started

Download the prebuilt installer for macOS, Windows or Linux from the official site or GitHub Releases. On macOS you can install via Homebrew:

brew tap poco-ai/agentero
brew install --cask agentero

Linux requires Ubuntu 22.04+ with webkit2gtk 4.1; the installation docs list the full dependencies. The CLI is a separate Homebrew formula:

brew tap poco-ai/agentero
brew install agentero
agentero --help

The README documents the desktop and CLI paths but not a complete first-run workflow (importing a paper, annotating it and asking an agent a question), so consult the online docs for the remaining steps.

Constraints

Running the app requires a desktop environment, and Linux users must satisfy the WebKit dependency. Agent connections depend on ACP-compatible agents that the user runs locally; external MCP clients are separate products. Remote access is done over SSH tunnels, and multi-device sync requires an S3-compatible bucket. The library is stored as plain PDF and Markdown files in standard folders rather than a private database, and Zotero databases can be imported with tags, notes and attachments preserved; BibTeX/BibLaTeX export is supported. The project is MIT-licensed; third-party components are listed in the notices file.

HotRepo's take

On the TypeScript daily list for 2026-11-08, Agentero ranked 13th with 1,951 stars and 124 forks, adding 175 stars in that day's window. That is the only recorded snapshot, so it supports a single-day observation rather than a trend. The feature set is concrete — structured extraction, ACP agents, MCP server, Zotero import and plain-file storage — which makes the project legible to researchers who already run local agents and want the library to be the shared context layer.

Project source and documentation

Written by AI from the project's README, updated . The project's own docs are the source of truth.