poco-ai/Agentero

What it solves

Agentero is a local-first research workbench designed to solve the problem of fragmented research contexts. It integrates PDF management, highlighted annotations, and AI-driven discussions into a single environment, preventing the loss of insights that typically occur when switching between separate tools like Zotero, Obsidian, and AI chat interfaces.

How it works

The platform uses a "Vault" system to keep work contexts local. It employs the Agent Client Protocol (ACP) to allow users to connect their own local agents (BYOA - Bring Your Own Agent), ensuring that the AI does not lock the user into a specific model. It also integrates with the Zotero ecosystem for importing libraries and supports the Model Context Protocol (MCP) to link with external clients like ChatGPT web.

Who it’s for

It is primarily for researchers and academics who need a unified space to manage literature, read PDFs with AI assistance, and maintain a structured knowledge base of notes and annotations.

Highlights

• Agent-Native Experience: Connect local agents via ACP to assist in retrieval, reading, and organizing workflows without model lock-in.
• Zotero Integration: Compatible with Zotero import methods and supports exporting to BibTeX/BibLaTeX for LaTeX writing.
• AI-Enhanced Reading: Features split-screen PDF reading with real-time AI summarization, translation, and the ability to parse figures, tables, and formulas within context.
• Extensible Capabilities: Supports the installation of community "Skills" to expand the agent's research and writing abilities.
• Local-First & Sync: Data is stored in local Vaults with optional S3-compatible cloud synchronization and SSH tunnel remote access.