unity-manual-mcp

unity-manual-mcp

Local MCP server that serves a local Unity documentation install to AI assistants, enabling full-text search and retrieval of Unity Manual and Script Reference pages through tools like search_unity_docs and get_page.

Category
访问服务器

README

Unity Manual MCP

Local MCP server (stdio) that serves a local Unity documentation install to your AI assistant.

The documentation folder is usually located in your Unity Editors install directory, under Data\Documentation\en.

  • Manual/ — how-to articles, organized in chapters (~3.5k pages)
  • ScriptReference/ — full scripting API reference, one page per class / property / method (~42k pages)

On first start the server builds a SQLite full-text index of every page (Mozilla-Readability-style extraction of the article body, boilerplate stripped, code examples preserved) and caches it in the OS temp dir. Subsequent starts reuse it; it is rebuilt automatically if the docs file count changes.

Requirements

  • uv
  • Python 3.10+ — declared in pyproject.toml (required by fastmcp 3.x). uv reuses any already-installed Python that meets this and only downloads a new one if no compatible interpreter exists.

Setup

  • Clone the repository
  • Open a terminal inside the cloned repository
  • Run
    uv sync
    

Creates .venv and installs the dependencies from pyproject.toml.

Run

uv run server.py "C:\Program Files\Unity\Hub\Editor\<ver>\Editor\Data\Documentation\en"

The en folder is auto-detected if you point at ...\Documentation or ...\Editor\Data instead.

Flags:

  • --selftest — build the index and run sample queries, then exit (no MCP server; good for verifying a fresh setup)
  • --rebuild — force rebuilding the index
  • --prune — drop stale cache entries (db file missing) from the manifest
  • UNITY_DOCS_PATH env var — alternative to the positional argument <br>

ℹ Run the server manually once after a fresh install to pre-index the docs and avoid waiting during first use

uv run server.py --selftest "C:\Program Files\Unity\Hub\Editor\<ver>\Editor\Data\Documentation\en"

Adding the MCP Server

This is a stdio server, so any client that can launch a local process works. Each client runs the same command:

uv run --directory "<repo>" server.py "<docs path>"

where <repo> is this repo's folder and <docs path> is your Unity docs en folder (auto-detected if you point at ...\Documentation or ...\Editor\Data instead). uv run creates and syncs the .venv environment automatically, so the client only needs uv on the PATH.

Claude Desktop

claude_desktop_config.json (substitute your own paths):

{
  "mcpServers": {
    "unity-manual": {
      "command": "uv",
      "args": ["run", "--directory", "<repo>", "server.py", "<docs path>"]
    }
  }
}

Claude Code

claude mcp add unity-manual -- uv run --directory "<repo>" server.py "<docs path>"

OpenCode

opencode.json in the project, or the global config (~/.config/opencode/opencode.json):

{
  "mcp": {
    "unity-manual": {
      "type": "local",
      "command": ["uv", "run", "--directory", "<repo>", "server.py", "<docs path>"],
      "enabled": true
    }
  }
}

Codex

codex mcp add unity-manual -- uv run --directory "<repo>" server.py "<docs path>"

Or edit ~/.codex/config.toml (shared with the ChatGPT desktop app and IDE extension; a project-scoped .codex/config.toml works in trusted projects):

[mcp_servers.unity-manual]
command = "uv"
args = ["run", "--directory", "<repo>", "server.py", "<docs path>"]

Tools

All read-only (readOnlyHint):

Tool Purpose
search_unity_docs Full-text search (all / manual / api)
get_page Read one page as clean text (path from search results)
list_sections Browse manual chapters, or pages inside a chapter
find_api Find a class / member API page by C# name

Index cache

%TEMP%\unity-manual-mcp\ holds one db file per editor (e.g. <hash>.db) plus a manifest.json mapping each docs path to its db file. Switching editors reuses the cached index; it is validated by HTML file count. Delete a db file (or use --rebuild) to force a rebuild. --selftest reports stale entries (db file missing) and orphaned entries (docs folder gone, db still there); --prune drops the stale ones.

Credits

Fully written by Qwen 3.8 27b 🙂

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选