DocsGPT is an open-source AI platform that enables organisations to deploy private AI agents, enterprise knowledge search, and automated data workflows on their own infrastructure or via a managed cloud service. Developed by Arc53 and released under the MIT licence, DocsGPT is designed for engineering teams, IT departments, and organisations that require full control over their data and AI stack. The platform connects to structured and unstructured data sources and returns answers grounded in the organisation's own content, with source citations included in every response. DocsGPT ingests content from a wide range of formats and sources, including PDF, DOCX, CSV, XLSX, HTML, Markdown, JSON, PPTX, audio files, GitHub repositories, web URLs, sitemaps, and relational databases. Key capabilities of DocsGPT include: - Private deployment: runs fully on-premises via Docker or Kubernetes, or as a managed cloud service at app.docsgpt.cloud, with no requirement to send data to third-party services - Multi-LLM support: connects to OpenAI, Anthropic, Google Gemini, and local inference engines including, allowing organisations to switch providers without re-architecting their deployment - Vector database compatibility: supports Qdrant, MongoDB, Elasticsearch, FAISS, Milvus, LanceDB, pgvector and more - AI agent builder: includes a visual workflow node editor for building multi-step agents that invoke tools, call external APIs, query databases, execute code, and trigger webhooks autonomously - Embeddable interfaces: provides React and HTML chat and search widgets for deployment on websites and documentation portals, alongside integrations with Slack, Telegram, Discord, and Chatwoot DocsGPT is used in production across more than 200 deployments by organisations including HM Government, PwC, Deloitte, and MIT. The project has accumulated over 18,000 stars on GitHub and is actively maintained with regular releases.