This project provides an autonomous AI research orchestration system designed to overcome the limitations of human cognitive architectures in scientific discovery. It targets researchers and power users seeking to leverage AI for deep scientific inquiry, enabling AI agents to autonomously conduct research from initial direction setting to experiment design without constant human intervention, thereby accelerating the pace of scientific breakthroughs.

How It Works

DARE employs a novel "pure-skill" architecture where all logic is encapsulated within over 800 Markdown files, executed by Claude Code. This approach eliminates traditional software infrastructure, requiring zero deployment or runtime dependencies beyond the execution environment. The system is organized into a four-layer military hierarchy (Campaign, Strategy, Tactic, SOP) for modularity and composability, allowing for flexible, non-linear research paths. It integrates with five external MCP servers (Semantic Scholar, Brave Search, AlphaXiv, Apify, Wiki Vault) for data acquisition and analysis, functioning as an "arsenal" of research methods rather than a fixed pipeline.

Quick Start & Requirements

• Primary install: Clone the repository (git clone https://github.com/yogsoth-ai/de-anthropocentric-research-engine.git) and run npm install.
• Prerequisites: Node.js (for npm), a Claude Code environment, and API keys for MCP servers (Semantic Scholar, Wiki Vault, Brave Search, Apify).
• Configuration: Copy mcp.example.json to .mcp.json and populate with API keys. Configure Claude Code's skills path to point to the cloned repository's skills directory.
• Execution: Invoke the entry point via /de-anthropocentric-research-engine for initial research direction crystallization or /executing-specs followed by a spec file path to run a pre-defined research plan.

Highlighted Details

• De-Anthropocentric Philosophy: Explicitly designed to remove human cognitive biases and limitations from the research process, arguing AI is better suited for complex, multi-domain scientific discovery.
• Arsenal, Not Pipeline: Research execution is dynamic, allowing agents to select, combine, and backtrack through various research strategies and tactics based on executable specifications, unlike rigid, pre-defined pipelines.
• Executable Research Specs: Generates machine-readable research plans that include progress tracking, quantified completion criteria, and explicit backtrack conditions, enabling autonomous execution and session recovery.
• Pure Markdown Skills: All operational logic is in Markdown files, facilitating universal composability, human readability, and instant modification without build steps or deployment.
• Four-Layer Hierarchy: Organizes over 800 skills into Campaign (research phases), Strategy (iteration engines), Tactic (workflow composition), and SOP (atomic operations) for structured autonomy.

Maintenance & Community

The project is developed by Pthahnix and is under active development. Near-term roadmap priorities include skill ablation for corpus refinement, advanced context engineering, cross-device session management, and developing an automated paper writing pipeline.

Licensing & Compatibility

The project is licensed under the Apache-2.0 license. This permissive license allows for commercial use and integration with closed-source projects.

Limitations & Caveats

The system's execution is fundamentally dependent on the Claude Code environment's ability to interpret and execute complex, multi-step instructions within Markdown files. The current skill corpus is deliberately over-complete, with plans for future refinement through ablation studies.