Which integrations are available for this server?

Allows AI agents to interact with a live Unreal Engine 5 editor session, including inspecting and modifying scene objects, reading output logs, searching C++ source files, triggering console commands, and scaffolding new C++ classes.

How do I use UE5 MCP Server?

1. Click on "Install Server". 2. Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state. 3. In the chat, type @ followed by the MCP server name and your instructions, e.g., "@UE5 MCP Server run console command stat unit" That's it! The server will respond to your query, and you can continue using it as needed. Here is a step-by-step guide with screenshots.

UE5 MCP Server

by Rolthar

Overview Schema Related Servers Score Discussions

TypeScript

Local

UE5 MCP Server

A Model Context Protocol server that bridges Claude AI to a live Unreal Engine 5 editor session. Claude can inspect and modify scene objects, read output logs, search C++ source files, trigger console commands, and scaffold new C++ classes — all from a natural language prompt.

How It Works

Claude ──JSON-RPC──► MCP Server (Node.js) ──HTTP──► UE5 Remote Control API (port 30010)
                          │
                          └──filesystem──► Source/*.cpp/h, Saved/Logs/*.log

Claude Desktop spawns the MCP server as a child process over stdio. When Claude decides to use a tool, it sends a JSON-RPC request; the server translates it into a UE5 Remote Control HTTP call (or a file system read/write) and returns the result.

Related MCP server: UE5 MCP Server

Prerequisites

Unreal Engine 5

Open your project in the UE5 Editor.
Go to Edit → Plugins, search for Remote Control API, enable it, and restart the editor.
Add the following to Config/DefaultEngine.ini:

[/Script/RemoteControlAPI.RemoteControlSettings]
bEnableRemoteControlHttp=True
RemoteControlHttpServerPort=30010
bRestrictServerToLocalHost=True

Security: Keep bRestrictServerToLocalHost=True. The Remote Control API has no authentication by default — exposing it to a network without protection is a serious risk.

Node.js

Node.js v20 or later is required.

Installation

cd C:/dev/ue5_MCP
npm install
npm run build

This compiles TypeScript to dist/. The runnable entry point is dist/index.js.

Configuration

Copy .env.example to .env and fill in your paths. The MCP server reads these as environment variables (injected via Claude Desktop config — see below).

Variable	Required	Default	Description
`UE5_RC_URL`	No	`http://127.0.0.1:30010`	Remote Control API base URL
`UE5_PROJECT_SOURCE`	Yes (for source tools)	—	Absolute path to your project's `Source/` directory
`UE5_LOG_PATH`	Yes (for log tool)	—	Absolute path to the `.log` file, e.g. `C:/…/Saved/Logs/MyProject.log`
`MCP_SECRET`	No	—	Adds `X-MCP-Secret` header to all Remote Control requests
`UE5_ALLOW_FILE_WRITE`	No	`false`	Must be exactly `"true"` to enable the file staging/write tools
`UE5_WRITE_EXTENSIONS`	No	`.cpp,.h`	Comma-separated allowed extensions for file writes
`UE5_STAGING_TTL_MS`	No	`600000`	Milliseconds before a staged batch expires (default: 10 minutes)

Connecting to Claude

Claude Code (CLI / VS Code)

Add a .mcp.json file in the root of the project you want the tools available in (or in the MCP server directory itself):

{
  "mcpServers": {
    "ue5": {
      "type": "stdio",
      "command": "node",
      "args": ["C:/dev/ue5_MCP/dist/index.js"],
      "env": {
        "UE5_RC_URL": "http://127.0.0.1:30010",
        "UE5_PROJECT_SOURCE": "C:\\Github\\MyProject\\Source",
        "UE5_LOG_PATH": "C:\\Github\\MyProject\\Saved\\Logs\\MyProject.log",
        "UE5_ALLOW_FILE_WRITE": "false"
      }
    }
  }
}

Alternatively, add it via the CLI:

claude mcp add --transport stdio \
  --env UE5_RC_URL=http://127.0.0.1:30010 \
  --env "UE5_PROJECT_SOURCE=C:\Github\MyProject\Source" \
  --env "UE5_LOG_PATH=C:\Github\MyProject\Saved\Logs\MyProject.log" \
  --env UE5_ALLOW_FILE_WRITE=false \
  ue5 -- node C:/dev/ue5_MCP/dist/index.js

When you start a new Claude Code session in a directory that has this .mcp.json, the 10 UE5 tools will be available automatically. You'll be prompted to approve the server on first use.

Claude Desktop (optional)

If you also want to use it with Claude Desktop, add to %APPDATA%\Claude\claude_desktop_config.json (Windows) or ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):

{
  "mcpServers": {
    "ue5": {
      "command": "node",
      "args": ["C:/dev/ue5_MCP/dist/index.js"],
      "env": {
        "UE5_RC_URL": "http://127.0.0.1:30010",
        "UE5_PROJECT_SOURCE": "C:\\Github\\MyProject\\Source",
        "UE5_LOG_PATH": "C:\\Github\\MyProject\\Saved\\Logs\\MyProject.log",
        "UE5_ALLOW_FILE_WRITE": "false"
      }
    }
  }
}

Restart Claude Desktop after editing. The hammer icon in the toolbar should show 10 tools when a conversation is active.

Available Tools

Remote Control (4 tools)

These tools communicate directly with the running UE5 editor via the Remote Control API.

`ue5_run_console_command`

Executes any Unreal console command in the live editor.

command: "stat fps"           → shows FPS overlay
command: "r.VSync 0"          → disables VSync
command: "stat unit"          → shows frame time breakdown
command: "showflag.Bloom 0"   → disables bloom

Blocked commands: quit, exit, open, cmd, exec, start, shell, servertravel, disconnect

`ue5_get_object_property`

Reads a property value from any UObject by its full actor path.

objectPath:   "/Game/Maps/TestMap.TestMap:PersistentLevel.PointLight_0"
propertyName: "Intensity"

`ue5_set_object_property`

Sets a property on any UObject.

objectPath:    "/Game/Maps/TestMap.TestMap:PersistentLevel.PointLight_0"
propertyName:  "Intensity"
propertyValue: 5000

`ue5_call_function`

Calls a Blueprint-callable UFUNCTION on any actor or component.

objectPath:   "/Game/Maps/TestMap.TestMap:PersistentLevel.MyActor_0"
functionName: "ResetState"
parameters:   {}

Source Search (2 tools)

These tools search and read files in your UE5_PROJECT_SOURCE directory. They never write anything.

`ue5_search_source`

Full-text regex search across .cpp and .h files with context lines.

pattern:      "AGameModeBase"
fileGlob:     "**/*.h"          (optional, default: **/*.{cpp,h})
maxResults:   30                (optional, default: 50)
contextLines: 3                 (optional, default: 2)

Example output:

=== MyGame/GameMode/MyGameMode.h:8 ===
   6:  #include "CoreMinimal.h"
   7:  #include "GameFramework/GameModeBase.h"
>> 8:  class MYGAME_API AMyGameMode : public AGameModeBase
   9:  {
  10:      GENERATED_BODY()

`ue5_read_file`

Returns the content of a specific source file, optionally scoped to a line range.

filePath:  "MyGame/GameMode/MyGameMode.cpp"
startLine: 20    (optional)
endLine:   50    (optional)

Log Reader (1 tool)

`ue5_read_log`

Reads the tail of the UE5 output log. Reads only the last N bytes to avoid loading gigabyte-sized session logs.

filter:    "Error"     (All | Error | Warning | Display, default: All)
maxLines:  200         (default: 100)
tailBytes: 1048576     (default: 512000 = 500 KB)

Use filter: "Error" after a failed compile to surface just the relevant errors.

File Writer (3 tools)

Requires UE5_ALLOW_FILE_WRITE=true in your environment. Disabled by default. Only .cpp and .h files are permitted. Paths are jailed to UE5_PROJECT_SOURCE.

The write workflow is a three-step process designed to give you a review gate before anything touches disk:

Step 1 — Stage

ue5_stage_files:
  files: [
    { path: "MyGame/MyActor.h",   content: "..." },
    { path: "MyGame/MyActor.cpp", content: "..." }
  ]

Files are held in memory. Nothing is written yet.

Step 2 — Preview

ue5_preview_staged

Returns the full content of every staged file. Review it before proceeding.

Step 3 — Commit

ue5_commit_files:
  triggerCompile: true

Writes all staged files to disk (creating directories as needed), clears the staging area, and optionally triggers a hot reload compile. Follow up with ue5_read_log filter: "Error" to check compile results.

Example Prompts

"Turn on GPU timing stats in the UE5 editor."
→ Calls ue5_run_console_command: "stat gpu"

"Set the intensity of DirectionalLight_0 to 3.14."
→ Calls ue5_set_object_property with the actor path and Intensity value

"Show me everywhere UHealthComponent is used in the project."
→ Calls ue5_search_source: pattern "UHealthComponent", maxResults 50

"My project just failed to compile. What are the errors?"
→ Calls ue5_read_log: filter "Error", maxLines 50

"Create a minimal AActor subclass called APickup with a USphereComponent."
→ Calls ue5_search_source to find conventions, then ue5_stage_files,
  ue5_preview_staged for your review, then ue5_commit_files with triggerCompile: true

Development

npm run dev       # Run directly with ts-node (no compile step)
npm run build     # Compile TypeScript → dist/
npm run start     # Run compiled output
npm run inspect   # Open MCP Inspector browser UI to test tools interactively

The MCP Inspector is the fastest way to test tools without Claude Desktop. It shows all registered tools with their schemas and lets you call them manually.

Security

Risk	Mitigation
Arbitrary console commands	Blocked prefix allowlist; `quit`, `open`, `exit` etc. are rejected
Remote Control API exposure	Keep `bRestrictServerToLocalHost=True` in `DefaultEngine.ini`
Path traversal in file reads	`../` patterns and escaped traversals are rejected before any `readFile`
Source file writes	Jailed to `UE5_PROJECT_SOURCE`; extension whitelist (`.cpp`, `.h` only); disabled by default
Log credential exposure	Logs may contain secrets from plugin init — redact before sharing raw log output
HTTP transport	If switching from stdio to HTTP/SSE transport, validate `MCP_SECRET` on every request

Project Structure

src/
  index.ts              Entry point — bootstraps MCP server, registers tools
  utils/
    ue5-client.ts       Axios wrapper for all Remote Control API HTTP calls
    log-parser.ts       Pure log line parsing/filtering (no I/O)
    path-guard.ts       Path traversal security guard (shared by read and write tools)
  tools/
    remote-control.ts   ue5_run_console_command, ue5_get/set_object_property, ue5_call_function
    source-search.ts    ue5_search_source, ue5_read_file
    log-reader.ts       ue5_read_log
    file-writer.ts      ue5_stage_files, ue5_preview_staged, ue5_commit_files
dist/                   Compiled JavaScript output (generated by npm run build)
.env.example            Template for environment variable configuration

This server cannot be installed

license - not found

quality - not tested

maintenance

How are these scores calculated?

Maintenance

–Maintainers

–Response time

–Release cycle

–Releases (12mo)

Commit activity

Resources

GitHub Repository

Need Help?

Related Servers

Unclaimed servers have limited discoverability.

Looking for Admin?

If you are the server author, to access and configure the admin panel.

Latest Blog Posts

Your AI Chatbot Just Exposed Your CEO's Salary to an Intern
By Om-Shree-0709 on July 2, 2026.
Agent Identity
MCP Security
OAuth Delegation
Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)
By Om-Shree-0709 on June 30, 2026.
Agentic Ai
Prompt Injection
WebAssembly
Lightport: Open-Sourcing Glama's AI Gateway
By punkpeye on April 27, 2026.
OpenAI
open source

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/Rolthar/Claude_UE5_MCP'

If you have feedback or need assistance with the MCP directory API, please join our Discord server