An open-source, API-first layer that provides Claude API-compatible endpoints over local models, so developers can build AI applications without rebuilding backend primitives or depending on cloud APIs. It requires a separately running OpenAI-compatible LLM server for inference and does not run models itself.
Project overview
It functions as a local backend underneath products rather than an app-first experience, supporting file ingestion with citations, structured database and CSV querying, MCP connectors, and custom tools.
Project type
Model Runtime
Use cases
Data Analysis
Deployment
Refer to project documentation
License
Apache-2.0
Best for
Developers, AI engineers, and enterprise teams building API-first applications who want to provide local backend functionality using local models.
Key capabilities
Supports file and artifact ingestion with retrieval augmented generation including citations.
Offers support for Model Context Protocol (MCP) in the API and remote MCP servers.
Includes built-in structured querying for databases and CSVs without relying solely on tools or code.
Provides the ability to configure and use custom tools.
Limitations and risks
Does not run models itself; it connects to OpenAI-compatible inference servers.
The UI is a demonstrator, not the core product.
Getting started
Setup requires installing PrivateGPT and separately setting up an OpenAI-compatible LLM server like Ollama. The first success path is to run 'private-gpt serve' and open http://localhost:8080/ui.