Workbench

Workbench is the built-in UI that ships with PrivateGPT. Once the server is running, open http://localhost:8080/ui — no extra setup needed.

It lets you try every API capability in a browser: chat with your documents, run web searches, query databases, invoke tools, and debug the underlying API calls in real time.

Workbench is a local demonstrator. It stores chats and settings in browser localStorage and has no access control — keep it local. For a production-ready platform, see Zylon.

Workbench does not guarantee exposure of every backend capability. Some features — including multimodality and other advanced or provider-specific capabilities — may be unavailable in the UI even when the API supports them.

PrivateGPT Workbench


Chat

The main area is a chat interface. Each conversation is saved locally in your browser and restored on reload. You can have multiple conversations and switch between them from the sidebar.

From the chat composer you can toggle which sources are active for that conversation — documents, databases, web, MCP servers, skills, or custom tools.


Context

The Context screen defines what the assistant can access. It has six sections:

SectionWhat it does
DocumentsUpload files and query them with retrieval and citations
DatabasesConnect SQL/NoSQL databases for natural-language queries
WebEnable live web search and URL fetch
MCPConnect MCP servers
SkillsEnable built-in skills like code execution
Custom ToolsDefine your own tools via JSON schema

API Debugger

The Debugger panel shows every request and response sent to PrivateGPT during your session. Use it to inspect the exact payload, verify which tools were included, or understand why the model responded the way it did.


Settings

Configure how Workbench connects to PrivateGPT: the API base URL, an optional bearer token, a global system prompt, and citation preferences.