RpgMakerMVUltimate MCP Server

RPG Maker MV Ultimate MCP Server

A Model Context Protocol server that lets an AI agent edit a real RPG Maker MV project on disk — database, maps, events, system — through 12 consolidated tools, validated against the actual engine so the output is coherent and playable.

Its headline feature is knowledge-driven map generation: instead of painting tiles procedurally, generate_map clones hand-authored reference maps from 106 bundled RTP templates (real multi-tile buildings, walls, furniture) and adapts them to your project's tilesets — falling back to procedural generation only for themes without a template.

The 101 fine-grained v4 tool names still work as call aliases (set RPGMV_LEGACY_TOOLS=1 to also advertise them).

Install

The package ships an executable, so no clone is required. Add it to your MCP client:

{
  "mcpServers": {
    "rpgmaker-mv": {
      "command": "npx",
      "args": ["-y", "rpgmaker-mv-mcp"],
      "env": {
        "RPGMAKER_PROJECT_PATH": "/path/to/your/RPGMakerMV/project"
      }
    }
  }
}

Works with Claude Desktop, Claude Code, opencode, and any MCP-compatible client. The server starts even without RPGMAKER_PROJECT_PATH — call set_project_path at runtime if you didn't set it.

From source

git clone https://github.com/DiegoLopez0208/RpgMakerMVUltimate-MCP
cd RpgMakerMVUltimate-MCP
npm install
npm run build
RPGMAKER_PROJECT_PATH=/path/to/your/project npm start

The 12 tools

Map generation (knowledge-driven)

generate_map defaults to mode: "procedural", which is smarter than the name suggests:

• For themes with a matching RTP reference template — town, village, dungeon, interior, castle, world, and more — it clones a hand-authored map from the 106 bundled templates, picking the closest size, so you get real 3D-looking buildings, walls and furniture instead of flat tile noise.
• For themes without a template (beach, swamp, desert, …) it generates procedurally (Perlin terrain, BSP dungeons, cellular caves).
• Same seed + params → the same map. Pass templateId to force a specific template, or useTemplate: false to force pure procedural.

Other modes: "blank" (empty canvas you paint with edit_map), "themed" (simple tile layout), "template" (instantiate one specific bundled map by templateId), "batch" (many maps at once), "duplicate" (copy an existing map).

{
  "tool": "generate_map",
  "arguments": { "mode": "procedural", "theme": "town", "name": "Riverbend", "width": 40, "height": 30 }
}

Combat themes (dungeon/cave/world/fortress/sewer/volcano) auto-wire random encounters from your existing troops, and town/village auto-create enterable house interiors with two-way warps. List available templates with get_project_context { detail: "templates" }.

Themes: forest town village castle dungeon cave beach desert swamp ruins interior snow harbor volcano sewer fortress magic_forest magic_interior space_interior space_exterior world

Offline map inspection

No API needed:

• query_map { view: "ascii", mapId } — render a map as a character grid with event markers (the cheapest way to "see" a layout and pick coordinates).
• query_map { view: "validate", mapId } — lint for invalid tile IDs, broken transfers, and missing event terminators.

Vision AI (optional)

analyze_image { mode: "ai" } sends a project image (tileset, sprite, screenshot, battler) to any OpenAI-compatible vision endpoint. It is disabled by default — nothing is sent anywhere unless you configure it. The other modes ("grid", "colors") and all other tools work fully offline.

# OpenAI
VISION_API_URL=https://api.openai.com VISION_API_KEY=sk-... VISION_MODEL=gpt-4o npm start
# Ollama (local, no key)
VISION_API_URL=http://localhost:11434 VISION_MODEL=llava npm start

Supported backends: OpenAI, Ollama, LocalAI, NVIDIA NIM, vLLM, LiteLLM, or any OpenAI-compatible proxy.

Agent Skill (recommended for any AI agent)

A portable Agent Skill teaches any model (Claude, DeepSeek, …) the correct, crash-free workflow — building maps with generate_map and adding content with manage_map_event presets instead of hand-painting tiles or guessing IDs. It lives at skill/rpgmaker-mv-mcp/SKILL.md.

# Claude Code / Claude.ai
npx degit DiegoLopez0208/RpgMakerMVUltimate-MCP/skill/rpgmaker-mv-mcp ~/.claude/skills/rpgmaker-mv-mcp
# opencode
npx degit DiegoLopez0208/RpgMakerMVUltimate-MCP/skill/rpgmaker-mv-mcp ~/.opencode/skills/rpgmaker-mv-mcp

The skill is also listed in awesome-claude-skills.

Knowledge base

Static JSON reference data in knowledge/, extracted from the RPG Maker MV corescript and the bundled maps:

Feedback & contributing

This server is actively developed and feedback is very welcome — bug reports, weird maps, missing tools, or ideas. Please open a GitHub Issue with what you asked the agent to do and what you got (an exported map JSON or a screenshot helps a lot).

Known limitations & roadmap

• Decoration/object semantics are best-effort; rare multi-tile objects may be placed as single tiles.
• Town and dungeon layouts keep improving — planned: central plaza/well landmark, houses in rows facing roads, fences/yards, richer road networks, more dungeon-room variety.
• Vision AI is optional and requires your own endpoint.

Development

npm install
npm run build      # tsc compile (+ copies knowledge/ into dist/)
npm test           # vitest
npm run dev        # tsx watch mode

Source: src/server.ts (tool handlers), src/toolDefinitionsV5.ts + src/v5Router.ts (the 12-tool v5 surface), src/tools/* (per-domain CRUD), src/utils/mapGenerator.ts (template cloning + procedural generation), knowledge/ (static reference data + bundled maps).

License

MIT