Where to Find PrivateGPT Documentation and Get Started

PrivateGPT's documentation lives at privategpt.dev, where the site's top navigation links directly to Blog, Docs, GitHub, and Join Discord. The Docs link is the main path for integration and API usage instructions; GitHub is where you inspect the source, star count, and derived projects; Discord is the community support channel. If you just want to see the product working before reading anything, the built-in /ui workbench is the fastest entry point.

The four entry points on privategpt.dev

The site reports 57K+ GitHub Stars, 7.5k+ derived projects, and 5k+ Discord members — useful signals if you're judging whether the project has enough surrounding activity to rely on.

What you're actually getting started with

PrivateGPT is described as an open-source API layer that turns local models into production AI applications, built by Zylon.ai. The important framing on the site: running a model locally is only the first step, and the API — not the workbench UI — is the actual product. The workbench exists for testing, demos, internal pilots, and inspecting API requests.

The API follows the Claude API model, so if you've worked with Claude-style message patterns, the shape will be familiar. The building blocks it ships:

• Messages API — streaming, async processing, token counting
• File & artifact ingestion — upload and process documents as context
• Retrieval with citations — RAG with source-grounded answers
• Structured data access — databases, CSVs, tabular data through the same layer
• Embeddings & orchestration — coordinate embeddings, retrieval, tools, and model calls behind one local API
• Custom tools & MCP — local MCPs or connected online providers such as n8n

Checking compatibility before you commit

Any tool that works with a local OpenAI-compatible provider will also work with PrivateGPT, per the site. That covers two directions:

Local model servers: Ollama, llama.cpp, vLLM, LocalAI, or your preferred OpenAI-compatible server.

Developer and automation tools: coding assistants such as Claude Code, OpenCode, VS Code, and Cline, plus automation platforms like n8n and internal systems.

If your stack already speaks the OpenAI-compatible protocol, the integration path is short. If it doesn't, check the Docs before assuming a drop-in fit.

A practical first-run order

1. Open Docs on privategpt.dev and read the integration and API sections.
2. Start a local model server (Ollama, llama.cpp, vLLM, or LocalAI) so there's something for the API layer to call.
3. Launch PrivateGPT and open the /ui workbench to confirm requests and responses behave as expected.
4. Move to the Messages API for real work — streaming, async, and token counting are where the production path begins.
5. Add file ingestion and retrieval with citations once basic calls are stable.
6. Join Discord if a step fails; the community is the support channel the site points to.

Common sticking points

• Treating the workbench as the product. The site is explicit that the API is the actual product and the UI is for testing and demos. Don't build your integration around the workbench.
• Skipping the local model server. PrivateGPT is an API layer over local models — it needs a running model server behind it.
• Assuming cloud independence is automatic. The value proposition is building without depending on cloud APIs, but custom tools and MCP can connect to online providers you trust (n8n is the example given). Decide deliberately which side of that line each integration sits on.

Where to go next

For the relationship between PrivateGPT and Zylon, the site's own comparison section ("PrivateGPT vs Zylon") is the place to look. For model and tool compatibility, the Docs plus the OpenAI-compatible rule above covers most cases. For everything else, GitHub for code and Discord for people.