Biomni

Report Abuse

Basic Information

Biomni is a general-purpose biomedical AI agent repository that provides software and tooling to autonomously execute diverse research tasks across biomedical subfields. It integrates large language model reasoning with retrieval-augmented planning and code-based execution to help scientists plan experiments, analyze data, and generate testable hypotheses. The project is distributed as a pip package and includes a scripted conda environment setup, a data lake that is automatically downloaded on first run, example notebooks and tutorials, and a no-code web interface. It supports multiple model providers via configuration, offers MCP (Model Context Protocol) support for external tool integration, and encourages community contributions to extend tools, datasets, and benchmarks. The README highlights security considerations because the agent can execute LLM-generated code with full system privileges, and it documents known package conflicts and release milestones.

Links

Categorization

App Details

Features
Autonomous biomedical task execution combining LLM reasoning, retrieval-augmented planning, and executable code. Multi-model support configurable via environment or .env, with explicit support for providers such as Anthropic, OpenAI, Azure, Gemini, Groq, Bedrock, Ollama, and custom endpoints. Packaged installation via conda environment and pip, plus an option to install directly from the repository. Built-in MCP support and example MCP server integrations to connect external tools and databases. A downloadable data lake used by the agent, example usages for CRISPR screen planning, scRNA-seq annotation, and ADMET prediction, plus notebooks and tutorials. A hosted no-code web UI and video demo. Contribution guides, a roadmap toward Biomni-E2, and explicit security and licensing notes are provided.
Use Cases
Biomni helps biomedical researchers increase productivity by automating complex research workflows that involve planning experiments, analyzing omics data, and generating hypotheses that are actionable and testable. Its retrieval-augmented planning enables context-aware recommendations and literature-informed decisions while code execution lets the agent run analyses and produce reproducible outputs. MCP integration allows connecting domain tools and databases so users can extend capabilities or incorporate institutional resources. The provided environment, tutorials, and web interface lower the barrier for adoption by non-developers, and the community-driven model invites contributions to add tools, datasets, and benchmarks, facilitating shared standards and collaborative development of biomedical actions.

Please fill the required fields*