BMTools is an open-source library for tool learning with large language models, developed by OpenBMB and published on GitHub with about 2,800 stars. It provides open implementations of the ChatGPT-Plugins idea, letting LLMs call external tools and APIs to complete real tasks.
BMTools gives developers a standard way to wrap tools, from web search and weather to databases and multimodal services, behind a unified interface that a language model can discover and invoke. It includes a growing collection of community-contributed tools, a framework for chaining tool calls under an LLM controller, and demos showing plugin-style assistants built on open models rather than closed platforms. The project helped define early open-source practice for tool-augmented LLMs and remains a reference for agent tooling design. It is Apache-2.0 licensed; the codebase was last updated in late 2023. The companion paper and tool list are linked from the repository, and its design later fed into OpenBMB’s broader agent research, including the ChatDev and MetaGPT ecosystems built by the same community.
BMTools is completely free and open source under the Apache-2.0 license. There are no paid tiers; you run it yourself and supply your own LLM access and any third-party API keys.
BMTools is for researchers and developers building tool-using LLM agents who want an open, extensible foundation instead of a proprietary plugin ecosystem.