AI Agents
bisheng
Enterprise LLM app platform: GenAI workflows, RAG, agents, model management, evaluation, fine-tuning.
★ 12.0k starsApache-2.0
View guide →Microsoft's R&D-Agent: LLM agents that automate data-driven research, Kaggle modelling and LLM fine-tuning.
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.
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 RD-Agent, then set your LLM configuration and scenario variables in the .env file, including the local data path for data-science runs.
Launch the scenario from the rdagent CLI, for example a data-science competition run; the agent proposes, implements, executes and iterates.
Watch run logs and results in the Streamlit UI (rdagent ui) or the web UI, and inspect the execution traces before trusting any output.
Agents read real-world material such as reports and papers, extract key formulas, features and models, and implement them in runnable code.
Because an LLM rarely implements an idea correctly at once, the framework evolves solutions by learning from execution feedback and knowledge.
A Kaggle-focused scenario automates feature engineering and model tuning, configured through .env variables such as the scenario class and data paths.
An autonomous fine-tuning scenario that handles data processing, training, evaluation and feedback-guided refinement for a base model you choose.
Applied to quantitative finance (including a multi-agent Quant variant) and medical model implementation, as copilot or autonomous agent.
A Streamlit UI for run logs (especially data-science) and a separate authenticated web UI for the Flask backend.
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-AgentSet 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 setupStart an autonomous data-science run on a competition, then monitor it with rdagent ui.
rdagent data_science --competition tabular-playground-series-dec-2021It is a Microsoft open-source project, with a published technical report (arXiv 2505.14738). Microsoft notes it is a research tool, not financial advice.
The code is MIT-licensed and free; you pay for the LLM calls and compute your runs consume.
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.
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.
Install the package, copy the .env configuration for the data-science scenario from the README, and run rdagent data_science on a sample competition.
Enterprise LLM app platform: GenAI workflows, RAG, agents, model management, evaluation, fine-tuning.
A low-code tool for building multi-agent LLM apps — chatbots, RAG and fine-tuning in one flow.
Research-backed lifelong memory for LLM agents, text and multimodal.
Chat with your data: an LLM agent that builds charts and business insights.
Agentic memory for LLM agents, from a published research project.
Microsoft's trainer for AI agents - reinforcement learning for almost any agent framework
The most-starred other AI Agents projects in the LibreAI catalogue — the realistic shortlist if RD-Agent is not the right fit.
A personal AI that really does things — on any OS, with your data staying yours.
An agentic skills framework & software development methodology that works.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
The agent that grows with you — a Python framework for adaptive AI agents.
DeepSeek's plugin-based agent harness where everything is a plugin.
Free and open — grab the source, star the repo, and start creating.