# NarshaADK

> Connect your AI to your Unreal Engine project. NarshaADK is an Unreal Engine MCP tool for C++, PDB symbols, and supported binary Blueprint data, with Unreal Editor closed for these reads.

Product facts reviewed: 2026-09-21. NarshaADK is the Fab product; NarshaMCP is its Unreal Engine MCP server, plugin folder, module, and executable. It connects to a separate AI client. Claude Code and Codex are tested clients.

## Research capability

- Supported source, symbol, and binary-asset reads do not launch Unreal Editor or commandlets. They can be used after project setup, a successful build, and indexing.
- PDB symbols supply compiled C++ relationships. Supported uncooked .uasset reads expose Blueprint events, variable defaults, node connections, and StateTree structures.
- The published examples connect C++ and Blueprint references and expose selected asset internals as structured MCP results. These are recorded results, not a live demo.
- Editor-closed research also exists in other tools. Compare direct project-file reads, previously built indexes, live Editor tools, and engine API references by the question being investigated. No comparative speed, token-use, or accuracy benchmark is claimed here.
- The demonstrated benefit is access to project information beyond source text. In the recorded Lyra cases, source search did not find FireDelayTimeSecs; binary analysis returned its saved default 0.1. The cross-reference result reports 2 Blueprint call sites. The StateTree result reports 9 states and 19 property bindings. These counts are scope-specific product outputs, not an independent accuracy evaluation.
- Supported reads remove an Editor or commandlet launch from the research workflow. These captures do not measure elapsed-time or memory savings. Some reads use metadata caches produced by binary parsing; unsaved Editor changes are not part of the files on disk.

## Requirements and limits

- Windows 64-bit; Unreal Engine 5.7 or 5.8 with a version-matched package; a C++ project; Visual Studio 2022 with Unreal C++ build tools. Blueprint-only projects are unsupported.
- PDB analysis needs one successful project build. Install engine debug symbols to include engine classes. Index loading or budget limits affect available results.
- Asset creation and editing, actor spawning, automation tests, and certain reads require a running Editor. See compatibility for per-operation exceptions.
- Cooked assets are unsupported. Blueprint data pins describe connections, not evaluated values. Latent, asynchronous, and delegate edges are approximated.
- Core analysis runs locally. Editor-closed operation does not promise an internet-disconnected AI client: clients may send supplied context under their provider policies. Optional anonymous telemetry is opt-in.
- Early Access: tool names, parameters, output formats, and coverage may change between releases. Pricing and license conditions are on Fab.

## Recorded evidence

The site's checked-in public README excerpts were recorded with NarshaMCP v0.13.7, Lyra Starter Game, UE 5.8.1, and Unreal Editor closed. They demonstrate those runs, not universal coverage. Excerpts preserve omission comments and are not complete unabridged runtime responses.

- [C++ to Blueprint references](https://narshaadk.ai/docs/technical-evidence/#cpp-blueprint): owning events, matched symbols, and qualified matches with confidence 0.75. This match score is not independently measured accuracy.
- [Inside a Blueprint](https://narshaadk.ai/docs/technical-evidence/#blueprint-data): events, variable defaults, and selected node connections from GA_Weapon_Fire.
- [StateTree structure](https://narshaadk.ai/docs/technical-evidence/#statetree): selected tasks and transitions, with omitted data marked; structural analysis, not runtime simulation.

## Documentation and purchasing

- [Product home](https://narshaadk.ai/)
- [Setup and client connection](https://narshaadk.ai/docs/getting-started/)
- [Research comparison and supported environments](https://narshaadk.ai/docs/compatibility/#research-fit)
- [Technical evidence](https://narshaadk.ai/docs/technical-evidence/)
- [Offline parsing benefits and measurement scope](https://narshaadk.ai/docs/technical-evidence/#offline-benefits)
- [Korean product home](https://narshaadk.ai/ko/)
- [Korean documentation](https://narshaadk.ai/ko/docs/getting-started/)
- [Public README and detailed comparison](https://github.com/Next-Stage-Inc/narshamcp-releases/)
- [Package releases](https://github.com/Next-Stage-Inc/narshamcp-releases/releases/)
- [Fab purchase, pricing, and support](https://www.fab.com/listings/c919281c-e2e6-4e81-8228-4e178cbc3e8d)
- [Privacy policy](https://github.com/Next-Stage-Inc/narshamcp-privacy/blob/main/PRIVACY_POLICY.md)