AI Agents
bisheng
Enterprise LLM app platform: GenAI workflows, RAG, agents, model management, evaluation, fine-tuning.
★ 12.0k starsApache-2.0
View guide →// agents · microsoft/RD-Agent
RD-Agent
RD-Agent is an open-source ai agents project by microsoft (microsoft/RD-Agent on GitHub). It has 14,866 GitHub stars and 1,953 forks, is released under the MIT license, and was last updated on GitHub on 6 October 2026.
Microsoft's R&D-Agent: LLM agents that automate data-driven research, Kaggle modelling and LLM fine-tuning.
★ 14.9k stars⑂ 2.0k forksMITupdated 2026-10
▦ visual — this project's README ships no embeddable screenshot,
so here's a placeholder. The guide below is complete regardless.
so here's a placeholder. The guide below is complete regardless.
// project facts
Project facts
- Repository
- microsoft/RD-Agent
- GitHub stars
- 14,866
- Forks
- 1,953
- License
- MIT
- Last updated on GitHub
- 6 October 2026
- Category
- AI Agents
- Listing
- Full install guide
- Official site
- Official site
// how it works
How it works
- Choose a scenario
Pick what to automate first: a Kaggle-style data-science task, finance or medical model implementation, or autonomous LLM fine-tuning with FT-Agent.
- Install and configure
Install RD-Agent, then set your LLM configuration and scenario variables in the .env file, including the local data path for data-science runs.
- Run the agent
Launch the scenario from the rdagent CLI, for example a data-science competition run; the agent proposes, implements, executes and iterates.
- Monitor and review results
Watch run logs and results in the Streamlit UI (rdagent ui) or the web UI, and inspect the execution traces before trusting any output.
// features
Why people love it
End-to-end R&D automation
Agents read real-world material such as reports and papers, extract key formulas, features and models, and implement them in runnable code.
Evolving, feedback-driven loop
Because an LLM rarely implements an idea correctly at once, the framework evolves solutions by learning from execution feedback and knowledge.
Data-science scenario
A Kaggle-focused scenario automates feature engineering and model tuning, configured through .env variables such as the scenario class and data paths.
FT-Agent for LLM fine-tuning
An autonomous fine-tuning scenario that handles data processing, training, evaluation and feedback-guided refinement for a base model you choose.
Finance and medical scenarios
Applied to quantitative finance (including a multi-agent Quant variant) and medical model implementation, as copilot or autonomous agent.
Monitoring UIs
A Streamlit UI for run logs (especially data-science) and a separate authenticated web UI for the Flask backend.
Advertisement
// run it
Run it yourself
- Clone the repository
Clone RD-Agent if you want the source and scenario examples; most users install the package instead.
git clone https://github.com/microsoft/RD-Agent.git cd RD-Agent - Configure the environment
Set the data-science scenario variables in .env as shown in the README before running.
dotenv set DS_LOCAL_DATA_PATH "$(pwd)/git_ignore_folder/ds_data" dotenv set DS_SCEN rdagent.scenarios.data_science.scen.KaggleScen # see README for the exact setup - Run a scenario
Start an autonomous data-science run on a competition, then monitor it with rdagent ui.
rdagent data_science --competition tabular-playground-series-dec-2021
// use cases
What you can build
- Automating a Kaggle competition entry: feature engineering, model search and iteration.
- Turning a quantitative-finance research report into implemented and backtested factors.
- Fine-tuning an LLM autonomously with FT-Agent against a benchmark description.
- Reproducing a paper's model from its text as a starting point for your own experiments.
Advertisement
// faq
Questions, answered
Who maintains RD-Agent?
It is a Microsoft open-source project, with a published technical report (arXiv 2505.14738). Microsoft notes it is a research tool, not financial advice.
Is it free to use?
The code is MIT-licensed and free; you pay for the LLM calls and compute your runs consume.
Does it need a paid API key?
Yes, in practice: RD-Agent is configured with an LLM through its .env settings (LiteLLM-style configuration), so you need your own model provider key or endpoint.
How does it compare with AutoML tools or a general coding agent?
AutoML tunes models within fixed search spaces, and general agents code to order; RD-Agent specifically automates the research loop - reading literature, proposing ideas, implementing and evolving them from feedback.
How do I get started?
Install the package, copy the .env configuration for the data-science scenario from the README, and run rdagent data_science on a sample competition.
// keep exploring
Similar tools
AI Agents
LazyLLM
A low-code tool for building multi-agent LLM apps — chatbots, RAG and fine-tuning in one flow.
★ 3.9k starsApache-2.0
View guide →AI Agents
SimpleMem
Research-backed lifelong memory for LLM agents, text and multimodal.
★ 3.8k starsMIT
View guide →Agents
Data-Analysis-Agent
Chat with your data: an LLM agent that builds charts and business insights.
★ 2.6k starsCustom
View guide →AI Agents
A-mem-sys
Agentic memory for LLM agents, from a published research project.
★ 401 starsMIT
View guide →AI Agents
agent-lightning
Microsoft's trainer for AI agents - reinforcement learning for almost any agent framework
★ 18.6k starsMIT
View guide →// alternatives
Alternatives in AI Agents
The most-starred other AI Agents projects in the LibreAI catalogue — the realistic shortlist if RD-Agent is not the right fit.
- openclaw★ 391.6k
A personal AI that really does things — on any OS, with your data staying yours.
- superpowers★ 294.9k
An agentic skills framework & software development methodology that works.
- ECC★ 275.2k
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
- hermes-agent★ 252.0k
The agent that grows with you — a Python framework for adaptive AI agents.
- deepseek-harness★ 245.5k
DeepSeek's plugin-based agent harness where everything is a plugin.
// more in agents
More in AI Agents
Nearby AI Agents projects in the catalogue, closest in stars to RD-Agent — keep exploring the shelf either side of it.
- botpress★ 14.9k
Open-source hub of integrations, SDK and CLI for building and deploying GPT/LLM chatbots and agents.
- unity-mcp★ 14.8k
Bridge AI assistants to the Unity Editor via MCP: manage assets, scenes, scripts and tasks.
- unstructured★ 15.5k
Turn messy documents into clean, structured data for language models.
- Agents★ 14.5k
A framework for building realtime voice AI agents 🤖🎙️📹
- OpenSandbox★ 15.7k
A secure, fast sandbox runtime where AI agents can execute safely.
- E2B★ 14.2k
Secure sandboxes where AI agents run real code.
Get RD-Agent
Free and open — grab the source, star the repo, and start creating.