Scribe is an MCP server built with Quarkus that exposes a single tool—executeKantraOperation—to generate YAML-based static analysis rules compatible with Konveyor/Kantra. Instead of hand-writing repetitive rule YAML, developers can prompt an LLM agent (Cursor, Claude Desktop, etc.) to produce valid Kantra rules from migration guides, tickets, or ad-hoc descriptions. The post covers Scribe's internal architecture (KantraTool, CommandRegistry, KantraCommand, Rule record, and polymorphic Condition types), provides three concrete rule examples (Java import migration, Go type reference for Kubernetes Ingress→Gateway API, and XML file content matching), explains how to connect it via mcp.json, and lists explicit limitations around rule validation, pattern noise, and JSON escaping.
Table of contents
Understanding the Scribe modelExample 1: Java import migrationExample 2: Go type reference (Kubernetes Ingress → Gateway API)Example 3: File content + XML (Struts / struts.xml )Running and connectingLimitations (explicit)Sort: