SkillAgentSearch skills...

enginewright

In-editor MCP server for Unreal Engine: strict, verified, undoable tool calls for AI agents. No Python/Node required. Early preview.

Install / Use

claude mcp add shanyup -- npx -y github:shanyup/enginewright

If the server publishes to npm under a different name, use that package instead — check the repo README.

About this skill
🔌

MCP Server

Model Context Protocol server

Quality Score

83/100

Supported Platforms

Claude Code
Claude Desktop
OpenAI Codex

Our assessment of enginewright

enginewright scores 83/100 on our quality scale, 2978th of 4,698 Development & Engineering skills we index.

Its MCP Server is 13 KB long, well organised into 14 sections with 2 code examples: a thorough specification that gives an agent plenty to work with.

It has 10 GitHub stars, so there is little community track record yet; judge it on its content.

Substance
30/30
Structure
18/20
Description
15/15
Adoption
4/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated 6 days ago, so enginewright is actively maintained.
  • Our last check on 2026-11-10 found the source still online.
  • It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
  • Its trust signals score 97/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

Safety scan

No issues found

Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.

AI review by kimi-k2.7-code on 2026-11-10. Automated pattern scan on 2026-11-09. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.

enginewright compared with similar skills

All 4 of these similar skills score higher than enginewright; compare them before choosing.

SkillScoreStarsUpdatedFormat
enginewright (this skill)by shanyup83106d agoMCP Server
taste-skillby Leonxlnx100116.8k8d agoCLAUDE.md
Agent-Reachby Panniantong100112.3k8d agoCLAUDE.md
rufloby ruvnet10081.9ktodayCLAUDE.md
headroomby headroomlabs-ai10079.8ktodayCLAUDE.md

Frequently asked questions

How do I install enginewright?
Run claude mcp add shanyup -- npx -y github:shanyup/enginewright. The install tabs above show the steps for each supported agent.
Which AI agents does enginewright work with?
It is written for Claude Code, Claude Desktop and OpenAI Codex, as a MCP Server file. Other agents that read the same format can often use it too.
Is enginewright safe to use?
Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It is MIT-licensed and scores 97/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is enginewright still maintained?
The repository was last updated 6 days ago, so enginewright is actively maintained.

Enginewright

Version Unreal Engine Platform License: MIT

An Unreal Engine editor plugin that runs a Model Context Protocol (MCP) server inside the editor, so AI coding agents (Claude Code, Codex, Cursor, VS Code, Hermes or any MCP client) can read and edit your project, play it and check what is on screen. Every write is validated first and read back after.

An agent builds a menu widget, checks it with widget_render, opens it in Play In Editor and clicks its buttons

Above: an agent calls Enginewright tools on a fresh UE 5.8 project. It creates a menu widget and checks each change with widget_render, starts Play In Editor, opens the menu, reads the buttons with ui_state, clicks Play, and gets COVERED when it tries a button that another widget draws over.

Early preview (v0.3). One maintainer, APIs may change before 1.0. Use it on a project under version control and review what it changes. See the v0.3.0 release notes.

Why

The focus is on edits you can trust:

  • Nothing extra to install. One native C++ editor plugin. No Python, Node, proxy process or Epic's experimental MCP plugin.
  • Strict validation. Unknown arguments, unknown properties or struct fields, wrong types and bad enum names are errors with a suggestion. They are never ignored.
  • All or nothing. If one value in a call is bad, nothing is written. Each write is one undo step and is read back and compared.
  • Rollback. batch runs several steps in one call; if a step fails, the earlier steps are undone.
  • Safe by default. Asset edits are refused while Play In Editor runs. Deleting an asset that is still used is refused, and the reply lists who uses it. A click on a button that another widget covers is refused. The server listens on localhost only and needs a per-project token.

Compared with Epic's MCP toolset in UE 5.8

UE 5.8 ships an experimental MCP toolset with a larger tool catalog than Enginewright. Enginewright makes a different trade-off. In our use of the 5.8 toolset, some writes reported success while a misspelled key was silently ignored, and some asset lookups gave misleading results during Play In Editor. Enginewright rejects those cases with an error. Epic's toolset exists only in 5.8; Enginewright does not depend on it.

Screenshots

| | | |---|---| | Fresh install: doctor reports a healthy editor | widget_render: a Widget Blueprint rendered to PNG without Play In Editor | | Fresh install. The prebuilt zip in a blank UE 5.8 project; doctor reports a healthy editor. | widget_render. A Widget Blueprint as a PNG at any resolution, no Play In Editor needed. | | ui_click refuses a covered button | game_run screenshots at two resolutions, put side by side by image_compose | | ui_click in Play In Editor. The click is checked like a real one: a button under another widget is reported as COVERED. | game_run + image_compose. The game started with a map URL at two resolutions, its screenshots collected and put side by side. |

Supported versions

| Unreal Engine | Platform | Status | |---|---|---| | 5.8 | Windows (Win64) | Tested (prebuilt zip and source) | | Other versions, Mac, Linux | — | Not tested yet |

Install (about 5 minutes)

These steps were run as written on a new, empty UE 5.8 project (no C++ code) with the v0.3.0 zip.

Prebuilt zip (Blueprint or C++ projects, UE 5.8 Win64)

  1. Download Enginewright-<version>-UE5.8-Win64.zip from Releases. Optional: compare its SHA-256 with the one in the release notes (Get-FileHash <zip> in PowerShell).

  2. Close the editor. Unzip into your project's Plugins/ folder (create the folder if it does not exist), so you get <YourProject>/Plugins/Enginewright/Enginewright.uplugin.

  3. Open the project. The plugin is enabled by default; nothing is compiled. The Output Log prints Enginewright: MCP server on http://127.0.0.1:3100/mcp (… tools). Client configs: <YourProject>/Saved/Enginewright/clients.

  4. Connect your client. Ready-to-paste configs (URL + token) are in <YourProject>/Saved/Enginewright/clients/:

    | Client | File | What to do | |---|---|---| | Claude Code | claude-code.txt | Run the command once in your project folder (or merge claude-code.mcp.json into .mcp.json). | | Codex | codex.toml | Append to ~/.codex/config.toml. | | Cursor | cursor.mcp.json | Merge into .cursor/mcp.json. | | VS Code | vscode.mcp.json | Merge into .vscode/mcp.json. | | Hermes | hermes.yaml | Merge under mcp_servers in your Hermes config.yaml. |

    The files hold your project's token: do not commit them (Saved/ is normally ignored). An agent that is already connected can print them with the client_config tool.

  5. Ask your agent to call doctor. It reports the engine version, the server, the open level and any problems, each with its fix. On a new project it usually warns that the editor slows down in the background (an Unreal default that makes Play In Editor tests slow); ask the agent to apply the fix it names (settings_set on EditorPerformanceSettings), then doctor reports healthy: true.

  6. Try it: ask the agent to create a Widget Blueprint with a button and show it with widget_render.

From source (C++ projects, or to build it yourself)

  1. Copy Plugins/Enginewright into your project's Plugins/ folder.
  2. Build your project's editor target (or open the .uproject and accept the rebuild prompt).
  3. Continue with step 3 above.

To use another port, set the ENGINEWRIGHT_PORT environment variable before starting the editor.

Tools

The client sees three meta tools: search (find a tool in plain words), describe (read its arguments) and call (run it). Behind them:

| Area | Tools | |---|---| | Objects | object_get, object_set | | Assets | asset_find, asset_exists, asset_references, asset_duplicate, asset_rename (updates referencers, removes the redirector, reports text and IDs that still name the old asset; dryRun to preview), asset_delete (refuses assets still in use), asset_save, asset_import (texture, sound, font, mesh, while the editor is open) | | Types and data | enum_create, struct_create, type_info, datatable_create, datatable_get, datatable_set_rows, datatable_remove_rows, datatable_rename_row | | Blueprints | blueprint_create, blueprint_info, blueprint_compile, blueprint_add_variables, blueprint_remove_variables, blueprint_add_component, blueprint_remove_component, blueprint_graph_read, blueprint_add_node, blueprint_connect, blueprint_set_pin, blueprint_remove_node | | Materials, meshes, textures | material_create, material_info, material_add_expression, material_connect, material_instance_create, material_instance_set, mesh_info, mesh_set_material, texture_info | | Enhanced Input | input_action_create, input_context_create, input_map, input_unmap, input_context_info | | Level | level_open, level_save, actor_find, actor_spawn, actor_set_transform, actor_delete, actor_bounds, world_trace, class_measure (size of a class or Blueprint without placing it) | | Play In Editor | pie_start (1-4 players; standalone, listen or client; waits until every player has a pawn), pie_stop, pie_state, pie_input (synthetic Enhanced Input action or key per player), screenshot (what a player sees, UI included, or the editor viewport), screenshot_ab (same view before and after a change, with the share of changed pixels) | | Run the game | game_run (the game in -game mode or a packaged exe, with a map URL that is passed unchanged and checked against the game's log; one run per resolution; verdict from log patterns; only the log lines you ask for; new PNG files collected with their real size), game_run_status (follow or stop a run), image_compose (PNG files side by side in one labelled image, read back) | | Multiplayer | pie_actors (what the server and each client see: count, net roles, owner, property values; can wait for replication), pie_net_emulation (lag, loss, duplicates, reordering; profiles off / average / bad) | | Readability | screen_coverage (how big an actor looks from a player's camera: pixels and share of the screen) | | UMG | widget_create, widget_tree, widget_get, widget_add (widget + properties + slot layout, checked), widget_set, widget_remove, widget_render (Widget Blueprint to PNG at any resolution, no Play In Editor) | | Runtime UI (UMG + CommonUI) | ui_state (widgets in a player's viewport, CommonUI activatable widgets and stacks, visible buttons with label, selected state and rectangle), ui_click (by name or label; refuses a covered button, or fires the click directly), ui_open / ui_close (show one screen in Play In Editor, or push it onto a running CommonUI stack), ui_input_type (switch a player to gamepad / mouse and keyboard / touch to check glyphs and focus), ui_input_audit (CommonUI input wiring: missing action rows, Enhanced Input actions that never fire, duplicate actions, no input data asset). CommonUI is optional: it is used when the project has it. | | Localization | stringtable_list, stringtable_get, stringtable_set (String Table assets), stringtable_compare (source vs translation: missing, extra, empty, untranslated, format argument differences), loc_culture (preview another language) | | Gameplay Tags | gameplay_tags_list, gameplay_tag_add (to a tag ini, strict names), gameplay_tag_remove (refused while assets use the tag) | | Build and tests | live_coding_compile, editor_restart (save, close, optionally build C++, reopen), build_status, automation_list, automation_run | | Checks | validate (compile Blueprints, data validation, broken references, map check), id_audit (IDs used but never defined, and the reverse), key_audit (hard-coded input keys in C++ and Blueprints), asset_search_text | | Editor | editor_state, console_command, log_read, log_wait (wait for a log line), settings_get, settings_set, doctor, client_config | | Multi-step | batch (atomic, rolls back on failure) |

Optional: compiling while the editor is closed

The editor locks its own DLLs, so a full C++ build of the editor target needs the editor closed. From inside the editor, editor_restart with build: true does that for you (save, close, build, reopen). For agents that work from a shell while the editor is closed, the zip also contains Plugins/Enginewright/Scripts/ewtool.py (Python 3, standard library): build waits for locked DLLs and names the process that holds them, and runs UnrealBuildTool detached, so the build finishes even when the calling process is killed. You do not need Python for anything else. See docs/EWTOOL.md.

Running the tests

The tests in tests/ are plain Python 3 scripts (standard library only; Python is needed for the tests, not for the plugin). They talk to a running editor. Use a separate, empty test project: the suites create and delete assets under /Game/_EWTest and add (then remove) actors in the open level. They create their own fixtures, so a new blank project works. Enable the CommonUI plugin in the test project for the CommonUI checks of the ui suite.

# bash / zsh
export ENGINEWRIGHT_TEST_PROJECT="/path/to/TestProject"
# PowerShell
$env:ENGINEWRIGHT_TES

Truncated for display — read the full file on GitHub.

Related Skills

View on GitHub
GitHub Stars10
CategoryDevelopment
Updated6d ago
Forks0

Languages

C++

Trust signals

97/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

1 info