PrivateGPT

PrivateGPT is a self-hosted API layer for developers building applications around local models. It requires a separate inference server, such as Ollama, llama.cpp, vLLM or LocalAI; it doesn't run models itself.

• Document retrieval: Ingest files and artifacts, retrieve relevant material, and return answers with source citations for retrieval-augmented generation (RAG).
• Structured data: Give applications access to databases, CSVs and tabular data through the same API.
• Agent tools: Combine code execution, custom tools and MCP connectors with web search and page fetching, which require internet access.
• Application API: Use Claude-style message patterns with streaming, asynchronous processing and token counting, while coordinating embeddings, retrieval and model calls.
• Existing assistants: Connect local models to Claude Code, OpenCode, VS Code and Cline, or use the backend in n8n automations.

The Python project is open source under Apache 2.0 and supports Docker deployment. Local workflows can operate offline and keep data within your environment. Connections to online providers send requests outside that environment.

Its built-in workbench is for testing, demos and inspecting requests; the main product is the developer API. Zylon is a separate enterprise platform from the same team that adds inference infrastructure, governance, user management and operational support.