UnrealClaude v1.4.5

Markdown chat rendering and trusted-mode script execution

Two community contributions: real Markdown rendering inside the in-editor Claude Assistant chat, and an opt-in setting that suppresses the per-script permission dialog for trusted MCP/agent-driven workflows.

Markdown rendering in the in-editor chat (#32)

The Claude Assistant chat panel now renders Markdown — code blocks, tables, lists, headings — instead of plain text. Adds a Plain Text toggle on the toolbar to switch back to the old selectable-text mode, and right-click-and-drag to scroll the chat panel up and down.

• Thanks to @baixuewen for the PR.
• Follow-up fix: GFM alignment markers (| ----: |) in table separators are now accepted, so any table with column alignment renders correctly instead of falling through to plain-paragraph rendering.

Auto-approve script execution setting (#33)

New setting under Project Settings → Plugins → Unreal Claude:

• Auto-approve script execution (default: OFF) — when ON, every Python / C++ / Console / Editor Utility script triggered through the MCP bridge or the in-editor chat runs immediately without showing the permission dialog. Each auto-approved script writes a LogUnrealClaude Log entry with its type and description, so the audit trail is preserved.
• Designed for trusted MCP/agent-driven workflows where the per-script confirmation becomes the dominant friction. Only enable on machines and projects where the connected client is trusted.
• Persisted to Config/DefaultEditor.ini under [/Script/UnrealClaude.UnrealClaudeSettings].
• Thanks to @jondewoo for the PR.

Pre-built Binaries

UnrealClaude-v1.4.5-Win64.zip ships the prebuilt Win64 editor module. Extract into:

<UE_5.7>/Engine/Plugins/Marketplace/UnrealClaude/

Linux and macOS users should build from source — see INSTALL_LINUX.md and INSTALL_MAC.md.

Links

• Installation
• MCP Bridge Setup
• Full Documentation
• Report Issues