AutoResearch Sibyl

v1.0.0Coding Agentsstable

Fully Autonomous AI Research System with Self-Evolution, built natively on Claude Code

ai-agentai-for-scienceai-scientistautomated-scienceautonomous-agents
Share:
247
Stars
0
Downloads
0
Weekly
0/5

What is AutoResearch Sibyl?

AutoResearch Sibyl is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to fully autonomous ai research system with self-evolution, built natively on claude code

Fully Autonomous AI Research System with Self-Evolution, built natively on Claude Code

This server falls under the Coding Agents category on MCPgee, the world's largest MCP server directory with 33,000+ servers.

Features

  • Fully Autonomous AI Research System with Self-Evolution, bui

Use Cases

Build fully autonomous AI research systems natively.
Enable self-evolving agents for scientific discovery.
Automate research workflows with AI.
LicenseMIT
Languagepython
Versionv1.0.0
UpdatedMay 21, 2026
Statushealthy
Maintenanceactive

Works with

ClaudeOpenAIwindowsmacoslinux

Installation

Manual Installation

npx autoresearch-sibylsystem

Configuration

Configuration Details

Config File

claude_desktop_config.json

Performance

Response Metrics

Response Time< 200ms
ThroughputMedium

Resource Usage

Memory UsageLow
CPU UsageLow

How to Set Up and Use AutoResearch Sibyl

AutoResearch SibylSystem is a fully autonomous AI research platform built natively on Claude Code that orchestrates multi-agent teams to conduct end-to-end scientific research. Given a research question, Sibyl autonomously performs literature reviews via arXiv, debates hypotheses across six specialized agents, schedules GPU experiments on remote SSH servers, generates LaTeX papers in NeurIPS format, and evolves its own prompts and workflows based on cross-project lessons learned. It is designed for AI researchers and ML engineers who want to delegate the entire research pipeline — from ideation to publication-ready manuscript — to an agentic system that runs unattended in a tmux session.

Prerequisites

  • Python 3.12 or newer and Node.js 18 or newer
  • Claude Code CLI installed and authenticated with a valid ANTHROPIC_API_KEY
  • tmux installed for persistent session management
  • SSH access to a GPU server for running experiments (optional but recommended)
  • Git to clone the repository
1

Clone the repository

Clone the AutoResearch SibylSystem repository to your local machine or a server with internet access.

git clone https://github.com/Sibyl-Research-Team/AutoResearch-SibylSystem.git
cd AutoResearch-SibylSystem
2

Set required environment variables

Export your Anthropic API key and enable the experimental multi-agent orchestration feature in Claude Code. These must be set before running the setup script.

export ANTHROPIC_API_KEY="sk-ant-..."
export CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1
3

Run the automated setup script

The setup.sh script installs all Python and Node.js dependencies, configures the required MCP servers (SSH and arXiv), and interactively collects your GPU server SSH credentials and project preferences.

chmod +x setup.sh && ./setup.sh
4

Configure your research project in config.yaml

Edit config.yaml in the project root to specify your GPU server alias, maximum GPU count, experiment language preference, and self-healing parameters.

# config.yaml
ssh_server: "default"
remote_base: "/home/user/sibyl_system"
max_gpus: 4
language: en
codex_enabled: false
self_heal_enabled: true
self_heal_interval_sec: 300
5

Launch Sibyl in a tmux session via Claude Code

Start a new tmux session for persistence, then launch Claude Code with the SibylSystem plugin directory. The --dangerously-skip-permissions flag allows Sibyl to autonomously execute shell commands, SSH calls, and file writes without per-operation prompts.

tmux new -s sibyl
claude --plugin-dir ./plugin --dangerously-skip-permissions
6

Initiate a research task

Once Claude Code is running inside tmux, describe your research question. Sibyl will autonomously plan the study, run a literature review, design experiments, execute them on your GPU server, and write the paper.

AutoResearch Sibyl Examples

Client configuration

The system runs through Claude Code CLI rather than a traditional MCP client config. A minimal claude_desktop_config.json is shown for reference if used as a standalone MCP server.

{
  "mcpServers": {
    "sibyl-ssh": {
      "command": "npx",
      "args": ["-y", "@fangjunjie/ssh-mcp-server"]
    },
    "sibyl-arxiv": {
      "command": "python3",
      "args": ["-m", "arxiv_mcp_server"]
    }
  }
}

Prompts to try

Research task prompts to give Sibyl once Claude Code is running in the plugin directory.

- "Conduct a systematic literature review on 'efficient fine-tuning methods for large language models' and summarize the top 20 papers."
- "Design and run an ablation study comparing LoRA ranks 4, 8, and 16 on the GLUE benchmark using our GPU cluster."
- "Write a NeurIPS-format paper on our recent experimental results stored in results/exp_001/."
- "Search arXiv for papers published in 2024 on 'mechanistic interpretability' and identify research gaps."

Troubleshooting AutoResearch Sibyl

Claude Code exits immediately with 'plugin directory not found'

Ensure you are running 'claude --plugin-dir ./plugin' from inside the cloned AutoResearch-SibylSystem directory. The ./plugin directory must exist; if setup.sh did not complete successfully, re-run it.

arXiv MCP server fails to start with ModuleNotFoundError

The arxiv_mcp_server Python package must be installed in your active Python environment. Run 'pip install arxiv-mcp-server' or activate the virtual environment created by setup.sh before launching.

Self-healing loop keeps restarting experiments without progress

Check the self-heal logs under logs/self_heal.log. A common cause is SSH credential expiry — refresh your SSH key or set a non-expiring key. Also verify the GPU server has sufficient disk space for experiment outputs.

Frequently Asked Questions about AutoResearch Sibyl

What is AutoResearch Sibyl?

AutoResearch Sibyl is a Model Context Protocol (MCP) server that fully autonomous ai research system with self-evolution, built natively on claude code It connects AI assistants to external tools and data sources through a standardized interface.

How do I install AutoResearch Sibyl?

Follow the installation instructions on the AutoResearch Sibyl GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.

Which AI clients work with AutoResearch Sibyl?

AutoResearch Sibyl works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.

Is AutoResearch Sibyl free to use?

Yes, AutoResearch Sibyl is open source and available under the MIT license. You can use it freely in both personal and commercial projects.

Browse More Coding Agents MCP Servers

Explore all coding agents servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.

Quick Config Preview

{ "mcpServers": { "autoresearch-sibylsystem": { "command": "npx", "args": ["-y", "autoresearch-sibylsystem"] } } }

Add this to your claude_desktop_config.json or .cursor/mcp.json

Read the full setup guide →

Ready to use AutoResearch Sibyl?

Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.

33,000+ ServersFree & Open SourceStep-by-Step Guides