EvoScientist

Why we included this project

EvoScientist runs a full scientific workflow as a coordinated team of six sub-agents handling planning, research, coding, debugging, analysis, and writing, from initial intake through verification. It is built on the DeepAgents framework and ships as a Python package with a CLI, a desktop WebUI, and chat integrations for Telegram, Slack, Feishu, and WeChat, so it can sit in a team's workflow as a persistent research assistant rather than a one-off script. The self-evolving layer is what sets it apart: the agent distills recurring patterns from its own memory into reusable skills, builds a knowledge graph that carries insights across sessions, and proposes new skills for a human to review. That human-on-the-loop design is the main draw for researchers and technical leads who want an opinionated, out-of-the-box agent for literature exploration, data analysis, and report writing, since it can run scheduled research on its own while keeping a person in the approval loop.