AIRAS - an open-source project for research automation

AIRAS is an open-source software for automated research, being developed to support the entire research workflow. It aims to integrate all of the necessary functions for automating research—from literature search and method generation to experimentation and paper writing—and is designed with the aim of enabling as many individuals and organizations as possible to contribute to open innovation in research automation.

Features of AIRAS include:

• Implemented as individual research processes

• Allows users to add their own original research processes

• Enables saving implemented code and executing it on large-scale computational resources through advanced integration with GitHub

• Supports not only fully automated research, but also interactive research in which users can collaborate with the system

Currently, it focuses on the automation of machine learning research.

Quick Start

Use AIRAS research tools (paper search, retrieval, hypothesis generation, experiment execution, paper writing) directly from an MCP client. No clone, no Docker — only uv is required.

Claude Code — plugin (recommended)

The plugin installs the MCP server and bundled research-workflow skills (auto-research for backend-LLM mode with API keys, auto-research-claude-code for key-free authoring by Claude Code itself) in one step:

/plugin marketplace add airas-org/airas
/plugin install airas@airas

Claude Code — MCP server only

claude mcp add airas -- uvx airas

Other MCP clients (mcp.json)

Add the following to your client's MCP configuration file (e.g. .mcp.json):

{
  "mcpServers": {
    "airas": {
      "command": "uvx",
      "args": ["airas"]
    }
  }
}

The server also exposes an MCP prompt, start_research, that walks any MCP client through the full research flow (in Claude Code: /mcp__airas__start_research).

See the MCP documentation for the full tool list and configuration options.

Related MCP server: pubmed-search-mcp

Roadmap

• Complete automation of machine learning research with code-based experimentation

• Autonomous research in Research in simulated robotic environments

• Autonomous research in Real-world robotics research.

• Laboratory Automation and autonomous research in various fields

Contact

We aim to build an operating system for automated research that enables humanity to discover scientific breakthroughs it has not yet reached.

If you are interested in this topic, please feel free to contact us at ulti4929@gmail.com.

About AutoRes

This OSS is developed as part of the AutoRes project.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contributions

By contributing to this project, you agree that your contributions are subject to the Contributor License Agreement (CLA) and may be used, modified, redistributed, and relicensed by the project owner, including for commercial, enterprise, and SaaS offerings.

See CLA.md for details.

Related projects

• AI-Research-SKILLs (Orchestra Research, MIT) — a library of library-specific ML engineering skills. AIRAS's experiment template installs it on code-generation runners so agents get framework-level guidance (fine-tuning, distributed training, inference); AIRAS's get_library_docs MCP tool complements it by pointing agents at each library's living documentation (llms.txt endpoints).

Citation

If you use AIRAS in your research, please cite as follows:

@software{airas2025,
  author = {Toma Tanaka, Takumi Matsuzawa, Yuki Yoshino, Ilya Horiguchi, Shiro Takagi, Ryutaro Yamauchi, Wataru Kumagai},
  title = {AIRAS},
  year = {2025},
  publisher = {GitHub},
  url = {https://github.com/airas-org/airas}
}