RD-Agent
Microsoft's autonomous R&D agent for iterative hypothesis generation, experimentation, and quantitative research automation.
RD-Agent
RD-Agent is Microsoft's framework for automating the research and development cycle. It models R&D as a loop of hypothesis → experiment → feedback, enabling agents to autonomously run data science competitions, build quant trading factors, perform medical research tasks, and implement ML ideas from papers.
Key Features
- Hypothesis-driven loop — generates testable hypotheses from a research goal, implements experiments, evaluates results, and refines
- Multi-domain — specialized scenarios for quantitative finance (factor research), data science (Kaggle-style), and general ML
- Code-first execution — experiments run as generated Python code in isolated Docker containers
- Paper-to-code — reads arXiv papers and implements the described method automatically
- Multi-agent collaboration — research agent and development agent work in tandem with shared memory
- Extensible scenarios — add custom R&D tasks by implementing the scenario interface
Quick Start
pip install rdagent
rdagent fin_quant # run quant factor research scenario
Install via ai-supply
npx ai-supply add microsoft-rd-agent-research-dev
Curated mirror of the open-source RD-Agent (MIT). Get it from the source.
Compromise signals — malicious or tampered code (leaked secrets, backdoors, a dropped executable) — reduce the score, and known dependency CVEs carry a bounded penalty (they warrant review but never QUARANTINE — update the dependency to clear). Other dangerous-by-capability traits are risk surface, expected for some capabilities. Every finding is mapped to its OWASP control below.
Findings mapped to the OWASP Top 10 for LLM Applications (2025) and the OWASP Machine Learning Security Top 10. Expand any flagged control for the exact findings — compromise reduces the score; expected/risk-surface do not, except a known CVE, which carries a small bounded penalty (high/critical → Review).
The same gate an agent runs before installing (POST /api/v1/trust/microsoft-rd-agent-research-dev/check). Click a policy:
Consume RD-Agent programmatically. Authenticate with an API key or session — see Authorize an agent.
# Agents: CHECK BEFORE YOU INSTALL (no auth) — score, grade, level, capability manifest
curl https://ai-supply.store/api/v1/trust/microsoft-rd-agent-research-dev
# Gate against your org policy (returns { pass, violations })
curl -X POST https://ai-supply.store/api/v1/trust/microsoft-rd-agent-research-dev/check \
-H "Content-Type: application/json" \
-d '{"minGrade":"B","denyPermissions":["shell"],"denyUnknownEgress":true}'
# CLI
npx ai-supply add microsoft-rd-agent-research-dev
# REST (install → download)
curl -X POST https://ai-supply.store/api/v1/listings/microsoft-rd-agent-research-dev/install \
-H "Authorization: Bearer $AIM_KEY"
# MCP tool
install_listing({ "slug": "microsoft-rd-agent-research-dev" })OpenAPI spec →Curated mirror — latest upstream source. See the repository for tagged releases.