QuPath-MCP: a state-aware interface for reproducible AI-agent workflows in digital pathology
Year: 2026
Type: software
Abstract: QuPath-MCP is a QuPath extension that embeds a Model Context Protocol (MCP) server, letting AI agents inspect, execute, verify and reproduce digital-pathology workflows in QuPath. It exposes 34 typed tools spanning image/project state, viewer perception, annotation, detection and measurement, script execution and workflow provenance. Every tool declares a machine-readable side-effect hint (read-only, write or destructive), destructive capabilities are individually gated, and a provenance layer records each executed step so a saved workflow can be replayed deterministically with no model in the loop. The extension is distributed as a single JAR that installs into QuPath 0.7.0 by drag-and-drop. It builds on QuPath (Bankhead et al., 2017).
Author: Lu, Zhaoqi
Institution: Hong Kong Polytechnic University
Language: English
Topic: AI in cancer detection
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7224849690