Key Takeaways
- EvoScientist is a self-evolving, multi-agent AI research system for end-to-end scientific discovery
- Ranked #1 on DeepResearch Bench II in March 2026; won Best Paper Award at ICAIS 2025
- 6 specialized sub-agents: Plan, Research, Code, Debug, Analyze, Write — forming a complete scientific workflow
- EvoMemory: persistent memory across sessions — context, preferences, and findings carry over
- Supports multi-provider LLM (Anthropic, OpenAI, Google, NVIDIA), 200+ built-in skills, MCP integration
- Multi-channel: CLI/TUI, Telegram, Slack, Feishu, WeChat, Discord, QQ
What Is EvoScientist?
EvoScientist is a self-evolving, multi-agent AI research system designed for end-to-end scientific discovery. It operates as a "human-on-the-loop" research buddy that co-evolves with you.
Core attributes:
- Category: Multi-agent AI research system
- Ranking: #1 on DeepResearch Bench II (March 2026)
- Awards: Best Paper Award, ICAIS 2025 (6 AI-generated papers accepted)
- Install:
uv tool install EvoScientist
The 6-Agent Team
| Agent | Role |
|---|---|
| Plan | Breaks down research goals into structured tasks |
| Research | Scans, summarizes, and synthesizes literature |
| Code | Implements experiments, data processing, and analysis |
| Debug | Diagnoses and fixes issues in code and pipelines |
| Analyze | Interprets results, generates visualizations, draws conclusions |
| Write | Produces publication-ready manuscripts in LaTeX/Markdown |
Complete workflow: intake → plan → execute → evaluate → write → verify.
Key Features
- EvoMemory: Persistent memory across sessions — context, preferences, and findings carry over
- Multi-Provider LLM: Switch between Anthropic, OpenAI, Google, and NVIDIA models via one config
- 200+ Built-in Skills: Predefined research skills, plus install more from GitHub
- MCP Integration: Connect to any MCP server for additional tools
- Multi-Channel: CLI/TUI hub with Telegram, Slack, Feishu, WeChat, Discord, QQ integrations
Quick Start
uv tool install EvoScientist
EvoSci onboard
EvoSciOptional channel extras:
uv pip install "EvoScientist[telegram]"
uv pip install "EvoScientist[all-channels]"FAQ
Q1: How does EvoScientist compare to AutoResearchClaw?
EvoScientist uses 6 specialized sub-agents with self-evolving capabilities and EvoMemory. AutoResearchClaw uses a 23-stage pipeline with multi-agent debate. EvoScientist is better for iterative research with persistent context; AutoResearchClaw is better for one-shot paper generation.
Q2: Does EvoScientist require a GPU?
No GPU required for the agent logic. GPU is beneficial if using local LLM inference via Ollama.
Q3: Can it produce publication-ready papers?
It produces high-quality drafts. Human review and refinement are still needed for publication quality.
Summary
EvoScientist is the top-ranked multi-agent AI research system on DeepResearch Bench II. With 6 specialized agents, persistent memory, and 200+ skills, it covers the full scientific workflow from literature review to paper writing. Best used as an iterative research companion that evolves with your project over time.