Quick Start Guide (v3.0)

1 Prerequisites

Before installing CLAUDIUS, make sure you have the following:

• ✓
  Unreal Engine 5.4 or later

  CLAUDIUS supports UE 5.4, 5.5, 5.6, and 5.7. Earlier versions are not supported.

• ✓
  A UE5 Project

  You'll need an existing project or create a new one. Any template works.

• ✓
  Epic Games Launcher

  CLAUDIUS is installed via the Epic Marketplace, so you'll need the launcher.

2 Installing CLAUDIUS

Step 2.1: Download from Epic Marketplace

After purchasing CLAUDIUS from the Epic Marketplace, open the Epic Games Launcher and navigate to your Library.

Navigate to Library > Vault in the Epic Games Launcher

Step 2.2: Add to Your Project

Find CLAUDIUS in your Vault and click "Add to Project". Select your target project from the dropdown.

Select your project and click "Add to Project"

Step 2.3: Enable the Plugin

Open your project in Unreal Engine. Go to Edit > Plugins, search for "CLAUDIUS", and ensure it's enabled. Restart the editor if prompted.

// CLAUDIUS is added to your .uproject file
{
  "Plugins": [
    {
      "Name": "Claudius",
      "Enabled": true
    }
  ]
}

Step 2.4: Verify the Plugin is Running

Open the Output Log (Window > Developer Tools > Output Log) and filter for Claudius. You should see something like this:

# Successful startup looks like this
LogClaudius: Claudius Subsystem Initialized
LogClaudius: HTTP server listening on 127.0.0.1:8080
LogClaudius: Watching .../Plugins/Claudius/claudius_request.json

The plugin also creates Plugins/Claudius/claudius_request.json on first run. This is the file you'll use to send commands.

3 Running the Setup Wizard

The wizard handles preset selection, auto-enable policy, and CLAUDE.md linking in one flow. Each step's response includes a next_call field telling you exactly what to send next, so it's easy to script.

Step 3.1: Start the Wizard

Open Plugins/Claudius/claudius_request.json and replace its contents with this:

{
  "id": "setup",
  "category": "config",
  "command": "run_setup",
  "parameters": {}
}

Save the file. Within 500 ms claudius_response.json updates with the wizard's first prompt:

{
  "command_id": "setup",
  "success": true,
  "output": {
    "prompt": {
      "step": "welcome",
      "question": "Welcome to CLAUDIUS v3.0...",
      "choices": ["yes", "skip"],
      "step_index": 0,
      "total_steps": 8,
      "next_call": "config.answer_setup { step: \"welcome\", answer: \"<choice>\" }"
    }
  }
}

Step 3.2: Walk Through the 8 Steps

Each subsequent step is sent the same way. The general shape:

{
  "id": "wizard",
  "category": "config",
  "command": "answer_setup",
  "parameters": {
    "step": "welcome",
    "answer": "yes"
  }
}

The eight steps and their choices:

1. welcome         → yes / skip
2. workflow        → fps / platformer / level_design / animation / open_world / general
3. experience      → bp_only / bp_cpp / engine_dev / mixed
4. ai_tool         → claude_code / cursor / codex / other
5. token_budget    → minimal / balanced / maximal
6. auto_enable     → ask / always / never
7. claude_md_link  → yes / skip
8. confirm         → yes / no

Step 3.3: Final Confirmation Response

When you answer yes to the confirm step, the response includes everything you need to know:

{
  "success": true,
  "output": {
    "result": "complete",
    "preset_applied": "fps",
    "auto_enable_policy": "ask",
    "enabled_commands": 105,
    "claudius_md_path": ".../Claudius/CLAUDIUS.md",
    "user_claude_md_path": ".../CLAUDE.md",
    "user_claude_md_state": "linked",
    "claude_md_linked_now": true,
    "estimated_tokens": 7345,
    "profile": { "workflow": "fps", ... }
  }
}

4 Smoke-Testing a Real Command

Send a real command to confirm the curated surface is wired up:

{
  "id": "list",
  "category": "level",
  "command": "list_all_actors",
  "parameters": { "max_results": 10 }
}

You should see "success": true with up to 10 actors in the response. If you instead see "command_disabled", the preset you picked doesn't include this command. Either let Claude surface the suggested enable call (default auto_enable_policy: ask), or load a broader preset:

{
  "id": "preset",
  "category": "config",
  "command": "load_preset",
  "parameters": { "preset": "everything" }
}

5 Linking Your CLAUDE.md

If you said yes to claude_md_link in the wizard, your project-root CLAUDE.md already has this single line in it:

<!-- CLAUDIUS: pulls in the auto-generated command index. Safe to move or wrap in your own text. -->
@Claudius/CLAUDIUS.md

That's the entire integration. We never touch the rest of your CLAUDE.md, even on regenerate. If you skipped this step or want to do it later, run:

{
  "id": "link",
  "category": "config",
  "command": "link_user_claude_md",
  "parameters": {}
}

This command is idempotent — safe to call twice. It returns state_before and state_after (missing, linked, or not_linked) so you can see exactly what changed.

6 Next Steps

You're up and running. Here's what to explore next, in order of likely value:

• Drive it from Claude Code: open your project folder in Claude Code and ask it to do real work. The Claude Code Integration tutorial covers v3 specifics.
• Understand the curation system: presets, enable/disable, and the ask-before-enable policy are all covered in Curation & Presets.
• Re-run the wizard any time: see The Setup Wizard for the full step-by-step reference, including what each choice changes under the hood.
• Add your own commands: drop a Python file with # @claudius headers into {Project}/Claudius/UserCommands/ — see Skill Packs & User Commands.
• Animation studio workflows: if you're doing shot-based work (technical explainers, forensic reconstructions, historical recreations), see The Animation Studio Preset.
• Switch to HTTP for sub-50 ms latency: the HTTP API Deep Dive covers the same JSON shape sent via POST.
• Browse the full catalog: open {Project}/Claudius/CLAUDIUS.md after the wizard runs — every command is listed with its enabled flag.