EvoScientist/EvoScientist

What it solves

EvoScientist is designed to automate the scientific research process, moving from a human-in-the-loop system to a "human-on-the-loop" paradigm. It enables AI scientists to autonomously explore, generate insights, and iteratively improve their own capabilities, acting as a research buddy that co-evolves with human researchers to internalize scientific judgment and scholarly taste.

How it works

The system employs a multi-agent team consisting of six specialized sub-agents (plan, research, code, debug, analyze, and write) that follow a structured scientific workflow: intake, planning, execution, evaluation, writing, and verification. It features a self-evolving memory system that distills observations into a growing knowledge graph and an "AutoSkills" mechanism that identifies recurring patterns in memory to propose new, reusable skills for the user to review.

Who it’s for

It is built for researchers and scientists who want an autonomous AI partner to handle the end-to-end research lifecycle, including data analysis, code execution, and scholarly writing.

Highlights

• Self-Evolving Capabilities: Automatically distills memory into a knowledge graph and generates new reusable skills from observed patterns.
• Multi-Agent Orchestration: A coordinated team of six agents specializing in different stages of the research pipeline.
• Flexible Interface: Accessible via Desktop WebUI, CLI/TUI, and multiple communication channels like Telegram, Slack, and WeChat.
• Extensible Tooling: Supports Model Context Protocol (MCP) servers and the installation of custom skills from GitHub.
• Autonomous Scheduling: Ability to run recurring research tasks on a cron-style schedule without constant human supervision.