tumourlove/monolith

What it solves

Monolith solves the "context window flood" problem when integrating AI assistants with Unreal Engine. Instead of registering thousands of individual tools—which consumes excessive tokens and confuses the AI—it uses a namespace-dispatch pattern. This allows an AI to discover available capabilities dynamically and only request detailed schemas when necessary, keeping the AI's focus on the actual problem rather than a massive tool catalogue.

How it works

Monolith implements the Model Context Protocol (MCP) to expose over 1,400 actions across 26 namespaces (such as Blueprints, Materials, Animation, and Niagara) to AI clients like Claude Code, Cursor, and Cline. It uses a central monolith_discover() tool for discovery and namespace-specific query tools (e.g., blueprint_query) to execute actions.

It includes a native C++ proxy (monolith_proxy.exe) to maintain AI sessions across editor restarts and a standalone query tool (monolith_query.exe) for offline access to engine source and project asset indices without needing to launch the Unreal Editor.

Who it’s for

Unreal Engine developers (UE 5.7/5.8) who use AI coding assistants to automate editor tasks, author assets, and navigate complex engine source code.

Highlights

• Massive Action Surface: Provides access to 1,400+ actions across domains including AI, Gameplay Ability System, Niagara, and Motion Matching.
• Reflection Intelligence: Features a layer for architectural decision-record harvesting, replication inspection, and UHT reflection-edge queries.
• C++ Authoring Ergonomics: Specialized lookups for include paths, signatures, and Build.cs dependencies to streamline C++ development.
• Live PIE Introspection: Ability to drive and inspect the editor during Play-In-Editor (PIE) sessions, including calling functions and setting control rotation.
• Engine Source Index: A deep, offline index of nearly 1 million engine symbols for rapid AI lookups.