Skip to main content
Glama
mcp-server-moegirl-wiki

mcp-server-moegirl-wiki

by YukiSakiko

mcp-server-moegirl-wiki

PyPI version License: MIT

English | 简体中文

MCP server for 萌娘百科 (Moegirl Wiki). Two tools: search entries and fetch a page's intro.

Requires the MCP Python SDK 2.x (mcp>=2,<3); see Compatibility if you are still on 0.2.x.

Install

uvx mcp-server-moegirl-wiki         # recommended
pipx run mcp-server-moegirl-wiki
pip install mcp-server-moegirl-wiki

Related MCP server: Internetsearch-mcp-server

Tools

search_moegirl(keyword: str, limit: int = 5)

Returns {"results": [{"title", "url", "snippet"}, ...], "count": int}.

get_page(title: str, max_length: int = 6250)

Returns {"title", "content", "url", "page_id"} — full plain-text rendering of the page (all sections, templates expanded, links resolved). Auto-follows redirects. content is truncated to max_length characters; default 6250 covers a typical character page in full, pass larger (e.g. 20000) for very long pages.

Environment variables

Variable

Default

Description

MOEGIRL_MAX_LENGTH

6250

Default max_length for get_page when the caller does not specify one. Useful for tuning per-deployment without forking the package. Invalid/non-positive values are ignored with a warning.

Configuration

Claude Desktop

{
  "mcpServers": {
    "moegirl-wiki": {
      "command": "uvx",
      "args": ["mcp-server-moegirl-wiki"]
    }
  }
}

MaiBot (config/bot_config.toml)

MaiBot's default config already has the [mcp] block — just replace the servers line:

servers = [{name = "moegirl", enabled = true, transport = "stdio", command = "uvx", args = ["mcp-server-moegirl-wiki"], env = {MOEGIRL_MAX_LENGTH = "6250"}, url = "", headers = {}, http_timeout_seconds = 30.0, read_timeout_seconds = 300.0, authorization = {mode = "none", bearer_token = ""}}]

Tweak MOEGIRL_MAX_LENGTH to taste (raise to 20000 for very long pages, lower to 1500 for quick lookups). If uvx is not on PATH, swap command/args for command = "pipx", args = ["run", "mcp-server-moegirl-wiki"].

Development

git clone https://github.com/XXXxx7258/mcp-server-moegirl-wiki
cd mcp-server-moegirl-wiki
uv sync --group dev
uv run pytest -m network
uv run ruff check .

Compatibility

Server version

MCP Python SDK

Notes

0.3.x

mcp>=2,<3

SDK 2.0 renamed FastMCP to mcp.server.mcpserver.MCPServer, so 0.2.x could not run on it. Hosts whose client still uses SDK 1.x work fine — the stdio handshake negotiates a shared protocol version (verified: 1.29.1 client ↔ 2.3.0 server).

0.2.x

mcp>=1.0

Crashes at startup on SDK 2.x (No module named 'mcp.server.fastmcp'). That is what 0.3.0 fixes; pin mcp<2 only if you must stay on 0.2.x.

Limitations

  • get_page returns rendered plain text (templates expanded, links resolved). Raw wikitext (prop=revisions, action=parse, ?action=raw, Special:Export) is blocked by Moegirl Wiki's site-wide ACL — empirically verified that bot-password login does not unlock it; only sysop/staff groups can call these actions.

  • Uses the mzh.moegirl.org.cn mirror.

  • Proxy variables are honoured via httpx's trust_env. Some NO_PROXY values are valid for curl but unparseable for httpx (e.g. a bare ::1 next to [::1]); in that case the server logs a warning and issues direct requests instead of failing every tool call.

License

MIT

Available Tools

2 tools
get_pageA

获取萌娘百科条目的纯文本全文(所有章节、模板已展开、链接已解析、自动跟随重定向、按 max_length 截断)。

Inputs: title (required), max_length.

search_moegirlA

搜索萌娘百科条目。先用 generator=search 全文索引(命中时同时返回摘要片段),0 命中或异常时退回 opensearch 标题模糊匹配以保命中率。

Inputs: keyword (required), limit.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.8/5.0

    Scored across 2 tools

    Disambiguation5/5

    The two tools have clearly distinct purposes: one searches for pages, the other retrieves the content of a specific page. There is no overlap or ambiguity.

    Naming Consistency5/5

    Both tool names follow a consistent verb_noun pattern (get_page, search_moegirl), using snake_case. The naming is clear and predictable.

    Tool Count3/5

    With only two tools, the server feels minimal for a wiki. While it covers basic search and retrieval, the scope is narrow; typically 4-8 tools would be more appropriate.

    Completeness2/5

    The tool surface has significant gaps for a wiki domain: missing operations like listing pages, retrieving sections, getting categories, or browsing history. Agents will likely fail to accomplish common wiki tasks.

    Maintenance

    ActivityMaintained
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers