Agent Laboratory is an open-source, end-to-end autonomous research workflow that uses LLM agents as research assistants: specialized agents carry your idea from literature review and planning through experimentation, code execution and report writing, with you as the human researcher directing the work. A companion framework, AgentRxiv, lets autonomous research agents upload, retrieve and build on each other’s papers so progress can accumulate across runs. The project has about 5,900 stars on GitHub and was last updated in 2025.
Doing research involves long stretches of mechanical work โ surveying papers, scaffolding experiments, running code, checking results and drafting write-ups. Agent Laboratory assigns each stage to specialized LLM-driven agents that collaborate in a pipeline: agents review the literature on your topic, formulate a concrete plan, implement and run experiments, analyze what happened and draft a research report. You stay in the loop as the principal investigator, supplying the idea and steering at key points rather than doing every mechanical step yourself. AgentRxiv extends the idea into a shared memory for agents: an arXiv-like server where agent-produced papers are stored and retrieved, so later runs can stand on earlier results instead of starting from zero. The project ships with a paper and multilingual documentation, and it is frequently cited in discussions of autonomous science.
Agent Laboratory is free and open source under the MIT license. Your only costs are the LLM API calls the agents make while researching and running experiments.
Agent Laboratory is for researchers, graduate students and R&D teams who want to accelerate the mechanical side of machine-learning research, and for AI researchers studying how far autonomous agents can go in doing real science.