Address Book

A local address book: search, add, edit and delete contacts, kept in a SQLite file next to the app. It runs as a native desktop window through pywebview, which starts Streamlit on a free loopback port and stops it when the window closes.

The same contacts are served to AI agents over an SSE MCP server that lives inside the app and is switched on from the MCP tab: host, port and an optional bearer token, plus ready-made configuration for Claude Code, Cursor, VS Code, Windsurf and other clients. Agents get five tools — list, get, create, update and delete — and a write from an agent makes the open window rerun at once, so the table updates without a click. A server left running is recorded and comes back up with the app.

GitHub: https://github.com/farukcan/addressbook_streamlit

Tech stack

• Python 3.12+, Streamlit 1.64, SQLite in WAL mode through the standard library
• MCP Python SDK 2.2 over SSE, served by uvicorn in a background thread
• pywebview desktop launcher, uv for dependencies, Pillow for the generated logo
• pytest, including Streamlit AppTest runs and an end-to-end MCP client over SSE

Screens