Tags: atomantic/UEMCP
Tags
Release v2.0.0: Enhanced Architecture & Production Readiness Major improvements to UEMCP with centralized version management, enhanced asset information, and improved reliability for production use. Key highlights: - Centralized version management system - Enhanced asset info with comprehensive details - Dynamic tool registry with fallback support - Improved error handling and type safety Full release notes: https://github.com/atomantic/UEMCP/blob/main/docs/release-notes/v2.0.0.md
fix: resolve E2E test 'Unknown command' errors and improve UE log che… …cking (#56) * fix: resolve E2E test 'Unknown command' errors and improve UE log checking - Updated Python listener command mappings to handle server-layer-only tools - Added proper error responses for MCP server layer tools (undo/redo/history/checkpoint/placement_validate) - Fixed E2E test UE log checking by reading log file directly instead of through MCP - Enhanced comprehensive test with intelligent actor cleanup (check existence before deletion) - Added server-layer tool awareness to prevent inappropriate direct Python calls 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <[email protected]> * fix: address code review feedback for file descriptor cleanup and magic numbers - Use try-finally block for proper file descriptor cleanup in UE log checking - Extract magic number 50000 to named constant LOG_BUFFER_SIZE - Ensures proper resource management even if file operations fail Addresses GitHub Copilot code review feedback on PR #56 * enhance: organize test actors in 'Test' folder for comprehensive cleanup - Updated test-socket-snapping.js to use unique timestamped actor names - Added 'Test/SocketTest' folder organization for socket snapping test actors - Updated test-comprehensive-mcp.js to use 'Test/Building/*' folder structure - Enhanced E2E test runner with comprehensive test actor cleanup - Cleanup now detects and removes all test actor patterns (Wall*, Foundation, WallSegment_*, etc.) - Prevents test actor accumulation across multiple E2E test runs Benefits: - No more duplicate named actors left behind - Easy identification of test artifacts in World Outliner - Comprehensive cleanup ensures clean project state - Better test isolation and repeatability 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <[email protected]> * reset calibration level * completely reset Calibration level * fix: resolve E2E test cleanup issues - no more leftover actors Root Cause Analysis: - MCPClient had 30-actor limit in Python listener, missing test actors beyond limit - SocketActor1/SocketActor2 not included in cleanup patterns - Test folder structure remained in World Outliner after tests Solutions Implemented: - Increased MCPClient level_actors limit from 30 to 100 actors - Added 'SocketActor' to testActorPatterns for comprehensive cleanup - Added Test folder structure cleanup via python_proxy - Enhanced cleanup logging with detailed actor deletion tracking Test Results: ✅ E2E cleanup now removes ALL 6 test actors (Wall*, Door*, Foundation, SocketActor*) ✅ Test folder structure automatically cleaned when empty ✅ Calibration level remains unchanged after test completion ✅ No manual cleanup required between test runs The E2E testing framework now maintains a completely clean project state, ensuring reliable and repeatable test execution without any leftover artifacts. 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <[email protected]> --------- Co-authored-by: Claude <[email protected]>
Release v0.9.0: Performance & Reliability UEMCP v0.9.0 delivers comprehensive performance optimizations making UEMCP production-ready for large-scale Unreal Engine automation. Key Features: - Batch operations framework (80-90% HTTP overhead reduction) - Memory management system for stable long-running sessions - Viewport optimization during bulk operations - Enterprise-grade security and validation enhancements This release transforms UEMCP from a development tool into a production-ready automation platform capable of handling enterprise-scale UE workflows. Full release notes: https://github.com/atomantic/UEMCP/blob/main/docs/release-notes/v0.9.0.md Complete v0.9.0 release including: - Fixed CI workflow for proper release artifact generation - Consistent version numbering across all package files - Comprehensive release process documentation
Release v0.8.0: Blueprint Creation & Undo/Redo System Major Features: - Blueprint creation tool for programmatic Blueprint class generation - Comprehensive undo/redo system with operation history tracking - Checkpoint system for batch operation management - Enhanced viewport camera controls with configurable angles Key Improvements: - Actor state capture for undo support - Operation history with configurable size limits - Tool mapping system for redo operations - Improved error handling with proper error responses - Enhanced ID generation using crypto.randomUUID() Technical Details: - Blueprint tools limited by UE Python API constraints - Tag-based metadata storage with documented migration path - Memory-managed operation history (100 operation limit) - Full TypeScript type safety maintained throughout Breaking Changes: - None Bug Fixes: - Fixed viewport_look_at hardcoded angle issue - Corrected Yaw calculation for camera positioning - Fixed error response formatting issues Documentation: - Added comprehensive warnings about tag metadata limitations - Documented future migration strategies for UE 5.5+ - Updated CLAUDE.md with viewport best practices
Release v0.7.0: Socket-Based Building & Enhanced Workflows Major Features: - Socket-based actor placement system - Batch spawn operations (4-5x performance) - Placement validation for modular building - FAB marketplace asset import - Enhanced asset information - 30 total MCP tools (up from 22) See release notes for full details.
Release v0.5.0 Major release with architectural improvements and documentation overhaul. Highlights: - Modularized server architecture for better maintainability - Reorganized Python plugin with proper package structure - Comprehensive documentation of learnings and workarounds - Fixed multiple viewport and import issues - Added actor_organize tool for World Outliner management See CHANGELOG.md for full details.
Release v0.4.0 - Fixed restart_listener to work reliably without crashes - Added mesh parameter to actor_modify tool - Enhanced help tool with better descriptions - Fixed available_commands in status endpoint - Improved port cleanup and thread management - Non-blocking restart via tick handler for thread safety
Release v0.3.0 - New `actor_duplicate` tool for copying actors with position offsets - Enhanced `level_actors` output with rotation, scale, and asset path - Improved MCP development philosophy - always improve tools instead of using python_proxy - 5x faster reconnection (5s health checks instead of 30s) - Duplicate any actor by name with optional position offset - Preserves rotation, scale, and folder organization - Auto-generates names or accepts custom names - Perfect for building multi-story structures - Now returns complete transform data (location, rotation, scale) - Includes asset path for static mesh actors - Case-insensitive filtering by name or class - Essential for debugging and understanding scene structure - Better error messages when actors are not found - Automatic folder path preservation during duplication - Type-safe interfaces for all new data fields - Comprehensive documentation comparing MCP tools vs python_proxy (85% code reduction\!) - Added test suite for all MCP tools - Updated CLAUDE.md with MCP tool development philosophy - Added comprehensive MCP tools comparison documentation - Improved troubleshooting guide for connection handling - Added workflow documentation for Claude Code users - Fixed level_actors filter parameter not working - Removed broken grid_snap tool (not implementable via Python API) - `level_actors` now returns additional fields (rotation, scale, assetPath) - The returned data structure is backwards compatible but contains more information
PreviousNext