Skip to main content

UnrealMCPHub

Central management platform for Unreal Engine MCP instances — bridge between AI agents and UE across the full development lifecycle.

Works with RemoteMCP (UE Editor plugin that exposes engine capabilities via MCP protocol). The Hub manages everything around it: install the plugin, compile, launch the editor, detect crashes, and transparently proxy all tool calls — so the AI agent only needs one MCP endpoint.

AI Agent ──► UnrealMCPHub ──► UE Editor + RemoteMCP plugin
              (this repo)       (blackplume233/UnrealRemoteMCP)

RemoteMCP runs inside UE Editor and provides 17+ tools across 6 domains (level, blueprint, umg, edgraph, behaviortree, slate). UnrealMCPHub runs outside UE Editor and can compile, launch, monitor, and proxy — even when UE is not running.

Features

  • Project setup — Configure .uproject once; engine auto-detected from registry
  • Build & launch — Compile via UBT, launch editor, wait for MCP readiness
  • Plugin install — One-click RemoteMCP installation (local copy or GitHub download)
  • Instance discovery — Scan ports to find running UE editors
  • UE tool proxy — ue_run_python, ue_call, ue_list_tools and domain dispatch
  • Crash resilience — Crash detection, report retrieval, restart flow
  • Multi-instance — Switch between multiple UE editors via use_editor
  • Session notes — Persist context for crash recovery
  • One-click overview — hub_status shows project, plugin, instances, and watcher state

Quick Install

# uv (recommended)
uv tool install unrealhub

# or pip
pip install unrealhub

Then add to .cursor/mcp.json:

{
  "mcpServers": {
    "unrealhub": {
      "command": "unrealhub",
      "args": ["serve"]
    }
  }
}

Option B: Zero-install (auto-fetch on first launch)

No download, no clone, no install. Paste this single block into .cursor/mcp.json and restart Cursor:

{
  "mcpServers": {
    "unrealhub": {
      "command": "uvx",
      "args": ["unrealhub", "serve"]
    }
  }
}

Requires uv (pip install uv or winget install astral-sh.uv). Cursor will auto-fetch the Hub on first launch.

Option C: Standalone executable (no Python needed)

Download the latest executable for your platform from GitHub Releases, then:

{
  "mcpServers": {
    "unrealhub": {
      "command": "/path/to/unrealhub",
      "args": ["serve"]
    }
  }
}
Platform File
Windows x64 unrealhub-windows-amd64.exe
macOS ARM unrealhub-macos-arm64
Linux x64 unrealhub-linux-amd64

Option D: Clone for development

git clone https://github.com/blackplume233/UnrealMCPHub.git
cd UnrealMCPHub
uv sync          # or: pip install -e .
{
  "mcpServers": {
    "unrealhub": {
      "command": "uv",
      "args": ["--directory", "/path/to/UnrealMCPHub", "run", "unrealhub", "serve"]
    }
  }
}

Quick Start

For AI Agents (Cursor / Claude / etc.)

This is the primary use case. Add UnrealMCPHub as an MCP server in your AI tool, then the agent handles everything through natural language.

Step 1: Install & Configure MCP (see Quick Install above)

HTTP mode (for shared / remote / multi-client use):

unrealhub serve --http --port 9422
{
  "mcpServers": {
    "unrealhub": {
      "url": "http://127.0.0.1:9422/mcp"
    }
  }
}

Step 2: Talk to the agent

Once configured, the AI agent has full access to all Hub tools. Example conversations:

You: "帮我编译并启动 UE 项目 D:/Projects/MyGame/MyGame.uproject"
Agent: [calls setup_project, compile_project, launch_editor automatically]

You: "在 UE 里创建一个蓝图 Actor"
Agent: [calls ue_get_dispatch to find blueprint tools, then ue_call_dispatch]

You: "UE 崩溃了怎么办"
Agent: [calls get_crash_report, shows crash info, offers restart_editor]

Step 3: What happens behind the scenes

Agent → Hub (setup_project)     # One-time project config, persisted
Agent → Hub (compile_project)   # Compiles via UBT, even without UE running
Agent → Hub (launch_editor)     # Starts UE, waits for MCP readiness
Agent → Hub (ue_run_python)     # Hub forwards to UE's RemoteMCP
                ↓
        UE Editor (port 8422)   # Executes Python, returns result

The agent only needs to know about the Hub — it never talks to UE directly.

AI Agent Decision Flow

Is project configured?
├── No  → Ask user for .uproject path → setup_project()
└── Yes → Is UE Editor online?
          ├── No  → Need plugin? → install_plugin() → compile_project() → launch_editor()
          └── Yes → Use ue_* tools directly
                    └── Crashed? → get_crash_report() → restart_editor()

For Humans (CLI)

unrealhub setup /path/to/MyProject.uproject   # Configure project
unrealhub serve                                # Start MCP server (stdio)
unrealhub serve --http --port 9422             # Start MCP server (HTTP)
unrealhub status                               # Show instance status
unrealhub discover                             # Discover running UE instances
unrealhub compile                              # Build active project
unrealhub launch                               # Launch editor

Tool Reference

Hub Management (always available, even without UE)

Tool Description
setup_project Configure project path (once, persisted to ~/.unrealhub)
get_project_config View current project configuration
hub_status One-stop overview of everything
compile_project Compile project via UBT
launch_editor Start UE Editor, wait for MCP readiness
restart_editor Restart a crashed editor
install_plugin Install RemoteMCP plugin
set_plugin_source Configure plugin download URL or local path
discover_instances Scan ports for running UE instances
use_editor Switch active UE instance (multi-instance)
get_crash_report Get crash details
add_note / get_notes Session notes for crash recovery context

UE Proxy Tools (forwarded to active UE instance)

Tool Description
ue_run_python Execute Python script in UE
ue_call Call any UE MCP tool by name
ue_list_tools List all tools from UE instance
ue_get_dispatch List domain tools (level, blueprint, umg, etc.)
ue_call_dispatch Call a domain-specific tool
ue_test_state Test connection to UE
ue_status Get active instance status

Architecture

┌─────────────────┐     stdio/HTTP      ┌──────────────────┐
│  Cursor / Agent │ ◄─────────────────► │  UnrealMCPHub    │
└─────────────────┘                     │  (FastMCP Server) │
                                        └────────┬─────────┘
                                                 │
                 ┌───────────────────────────────┼───────────────────────────────┐
                 │                               │                               │
                 ▼                               ▼                               ▼
         ┌───────────────┐              ┌───────────────┐              ┌───────────────┐
         │ Project Mgmt  │              │ Lifecycle     │              │ UE Proxy      │
         │ setup_project │              │ compile       │              │ ue_run_python  │
         │ install_plugin│              │ launch/restart│              │ ue_call        │
         │ hub_status    │              │ discover      │              │ ue_list_tools  │
         └───────────────┘              └───────┬───────┘              └───────┬───────┘
                 │                              │                              │
                 ▼                              │  Streamable HTTP             │
         ┌───────────────┐              ┌──────▼──────────────────────────────▼────────┐
         │ ~/.unrealhub  │              │  UE Editor + RemoteMCP (port 8422)            │
         │ config.json   │              │  17+ tools: run_python, get_dispatch, etc.    │
         │ state.json    │              │  6 domains: level, blueprint, umg, edgraph,   │
         └───────────────┘              │             behaviortree, slate               │
                                        └──────────────────────────────────────────────┘

Development

git clone https://github.com/blackplume233/UnrealMCPHub.git
cd UnrealMCPHub
uv sync --extra dev      # Install with dev dependencies
uv run pytest tests/ -v  # Run tests (129 tests)

Requirements

  • Python >= 3.11
  • Unreal Engine 5.x
  • RemoteMCP — UE Editor plugin (Hub can auto-install it via install_plugin)

License

MIT

Metadata

Release files for unrealhub 0.3.1

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for unrealhub 0.3.1
File Size Uploaded Provenance Actions
unrealhub-0.3.1.tar.gz 76.9 kB With provenance Details for unrealhub-0.3.1.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for unrealhub 0.3.1
File Interpreter ABI Platform Actions
unrealhub-0.3.1-py3-none-any.whl Python 3 none any Details for unrealhub-0.3.1-py3-none-any.whl

Total release size: 127.0 kB

Release files / unrealhub-0.3.1.tar.gz

Download URL unrealhub-0.3.1.tar.gz
Size 76.9 kB
Tags Source
SHA-256 checksum
How to use checksums
69987086f7a34ea7e7383fe9f6af9a57f10a89a3df9f09d8e386db545f6a6274
BLAKE2b-256 checksum
How to use checksums
8f2b5406cc2a7d1b4e76e16c7e026873d071cd492a45a42493b3eb375118e264
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/6.1.0 CPython/3.13.7

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Mar 23, 2026.

Transparency log

Release files / unrealhub-0.3.1-py3-none-any.whl

Download URL unrealhub-0.3.1-py3-none-any.whl
Size 50.1 kB
Tags Python 3
SHA-256 checksum
How to use checksums
bd4f0048c8e88413f6044b38dc299afb94154d750582c14a706028d2ab204e7f
BLAKE2b-256 checksum
How to use checksums
5642c6aa1766c520df2f0deb733b62065d6dff3bc97c6b58fd38b50e1d2bd8f3
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/6.1.0 CPython/3.13.7

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Mar 23, 2026.

Transparency log

Security for unrealhub 0.3.1

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

All 2 files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for unrealhub-0.3.1.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Mar 23, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page