DB-GPT is an open-source agentic AI data assistant for building the next generation of AI-plus-data products. It has earned more than 20,000 stars on GitHub, is under active development with updates in October 2026, and is released under the MIT license.
DB-GPT lets people talk to their data. Connect it to databases and documents, and users can ask questions in natural language while agents plan the work: writing and checking SQL, retrieving supporting documents through RAG, and presenting answers as text, tables, and charts. The project puts privacy first, supporting fully local deployment with open models so sensitive enterprise data never has to leave the building. Around the core sits an agent framework for composing multi-step data workflows and a growing set of ready-made data applications.
DB-GPT has grown into a small ecosystem rather than a single app. Its agentic workflow language lets developers compose reusable data-processing agents, while companion projects focus on fine-tuning models for text-to-SQL and evaluating generated queries. Knowledge bases can be organized into spaces that mix database schemas, documents, and business glossaries, giving agents the context they need to answer domain questions accurately. Because the stack is model-neutral and deploys on-premise, it is popular with organizations in finance, government, and industry that want AI over their data without surrendering control of it.
DB-GPT is completely free and open source under the MIT license. Organizations can self-host it without license fees, paying only for their own infrastructure and model usage.
DB-GPT is for data teams, analysts, and enterprises that want conversational access to databases, automated analysis, and AI data assistants, especially where data privacy rules out sending information to external cloud services.