Fonto Docs

io.github.DrRataplan由 GitHub 帳號 DrRataplan 發布v1.8.1更新於 Oct 24, 2026

Fonto (FontoXML) documentation for AI tools. Converts DITA XML to Markdown on demand.

已驗證Streamable HTTP可網頁執行知識與記憶開發者工具

概覽

AI 產生的概覽

遠端 MCP 伺服器,讓助理以 Markdown 形式搜尋並閱讀 Fonto(FontoXML)文件及其 API 參考。

可以做什麼
  • 依關鍵字全文搜尋 Fonto 文件,回傳附標題、描述與 slug 的相關度排序結果
  • 依 slug 取得文件頁面的完整 Markdown 內容
  • 依精確名稱、呼叫寫法或匯入路徑查詢 API 符號,一次呼叫即回傳其文件頁
  • 依標題、產品或階層關鍵字篩選頁面清單,或讀取完整目錄資源
  • 透過頁面資源範本直接依 slug 存取任一頁面
適用情境
適合在進行 Fonto XML / FontoXML 編輯器開發時使用,讓助理根據即時文件回答 API、設定、元件或升級指南相關問題,而不是憑空推測。
執行需求
需要一個遠端 streamable HTTP 端點,無需本機安裝、帳號、API 金鑰或 OAuth。用戶端需支援遠端 HTTP MCP 伺服器或自訂連接器。伺服器本身需存取 Fonto 文件網站,因此其端需要網路連線。
安裝前請注意
僅提供唯讀查詢,不索取任何憑證,也不會寫回資料。查詢詞與頁面請求會送到第三方託管的伺服器,該伺服器再從 Fonto 文件網站擷取頁面,因此搜尋詞與所請求的 slug 對該營運方可見。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 Fonto Docs,將其新增到工作區。
  2. 為需要使用其工具的對話啟用該服務。

Web executable,透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。

交給代理程式安裝

把這段提示詞貼上到 Claude Code、Codex、Cursor 等代理程式,或 SourceWeft 對話中。代理程式會讀取該伺服器的安裝指南,先向你展示發布者、端點和登入方式,經你同意後再把它新增到你的客戶端。金鑰不會寫進任何命令。

閱讀 https://sourceweft.com/mcp/io.github.DrRataplan%2Ffonto-docs/install.md,並按其中說明把這個 MCP 伺服器接入我的客戶端。修改任何設定之前,先告訴我它的發布者、端點和登入方式,等我確認。不要把 API 金鑰寫進命令或我會提交的檔案;問我金鑰應從哪裡取得。

檢視代理程式所依據的安裝指南

其他 MCP 客戶端

把它新增到你客戶端的 mcpServers 設定中。

{
  "mcpServers": {
    "fonto-docs": {
      "type": "http",
      "url": "https://fonto-docs.elliat.nl/mcp"
    }
  }
}

README

fonto-docs-mcp

[Deploy to Cloud Run] [MIT License]

An MCP server that makes the Fonto XML documentation accessible to AI tools like Claude Code, Cursor, and Claude Desktop. Live at fonto-docs.elliat.nl.

The Fonto docs are rendered by a JavaScript SPA, which makes them impossible for AI to read directly. This server fetches the underlying XML and converts it to clean, readable Markdown on demand.

What is MCP?

MCP (Model Context Protocol) is a standard way to give AI assistants access to external tools. Once you connect this server to your AI tool, it gains access to these tools and resources:

ToolWhat it does
search_fonto_docsSearch by keyword — returns matching pages with titles, descriptions, and slugs
get_fonto_pageFetch the full content of a page by its slug
lookup_apiResolve an API name (e.g. documentsManager, createIconWidget) straight to its page in one call
list_pagesList all pages matching a keyword, with full section hierarchy — useful for discovery
ResourceWhat it contains
fonto://catalogAll ~2000 pages with real titles, product grouping, and ancestry paths
fonto://page/{slug}Resource template — address any page directly by its slug

You can then ask things like "How does addDocumentChangeCallback work?" and the AI will look it up in the live Fonto docs.

Connect to your AI tool

The server is already running at https://fonto-docs.elliat.nl/mcp — you just need to point your tool at it.

Claude Code (CLI)

bash
claude mcp add --transport http fonto-docs https://fonto-docs.elliat.nl/mcp

Cursor

Add to .cursor/mcp.json in your project (or ~/.cursor/mcp.json globally):

json
{  "mcpServers": {    "fonto-docs": {      "type": "http",      "url": "https://fonto-docs.elliat.nl/mcp"    }  }}

Claude Desktop / claude.ai

Add it as a custom connector: go to Customize → Connectors, click + → Add custom connector, and paste https://fonto-docs.elliat.nl/mcp. No OAuth settings are needed. Connectors added this way are available in both Claude Desktop and claude.ai. See Get started with custom connectors using remote MCP.

(claude_desktop_config.json only launches local stdio servers, so it can't be used for this server.)

Mistral AI / Vibe

The easiest way is through the Connectors page in the Mistral console:

  1. Go to the Connectors page and click + Add Connector.
  2. Open the Custom MCP Connector tab.
  3. Enter a name (e.g. fonto-docs) and the server URL https://fonto-docs.elliat.nl/mcp, then click Connect.

No authentication is needed. Vibe automatically discovers connectors registered this way, so the tools are available in every Vibe session where you're signed in with your Mistral account — no local configuration required.

Alternatively, add the server to your local Vibe configuration by editing ~/.vibe/config.toml:

toml
[[mcp_servers]]name = "fonto-docs"transport = "streamable-http"url = "https://fonto-docs.elliat.nl/mcp"

No auth block is needed — the server requires no authentication.

See the MCP Connectors documentation for more details.

Usage examples

Once connected, ask your AI assistant:

  • "Search the Fonto docs for documentsManager"
  • "Get the Fonto docs page for clearundostackfordocument-f0187fade723"
  • "How does addDocumentChangeCallback work according to the Fonto docs?"
  • "List all pages in the configure section"
  • "What upgrade guides are available?"

HTTP API

The server also exposes a plain HTTP API if you want to use it without MCP:

  • GET /search?q={query} — search pages by keyword
  • GET /page/{slug} — fetch a page as Markdown
  • GET /api/{name} — fetch the page for an API symbol by exact name, as Markdown (404 with suggestions if none)
  • GET /catalog — full page catalog grouped by section; add ?section={keyword} to filter

How it works

The Fonto documentation site stores its content as XML at predictable URLs under /static/xml/. This server fetches those XML files directly and converts them to Markdown, bypassing the JavaScript rendering. Page content is cached in-process for 10 minutes to absorb repeated lookups in the same session; after that, get_fonto_page fetches it live from documentation.fontoxml.com again. The page catalog (used by list_pages and fonto://catalog) is fetched once from the Fonto search index on first use and held in memory for the lifetime of the process.

Self-hosting

bash
npm installnpm start        # runs on port 8080 by defaultPORT=3000 npm start

Contributing

PRs welcome. The XML-to-Markdown conversion in src/fonto.js handles two formats:

  • DITA guide pages — <topic>, <body>, <section> structure
  • API reference pages — custom <type>, <members>, <description> structure

If you find pages that don't convert well, open an issue with the slug.

License

MIT

來源:README.md,提交 ebe2b49

工具

4

search_fonto_docs

Search Fonto Documentation

Search the Fonto XML documentation using full-text search. Returns results ranked by relevance with titles, descriptions, and slugs. Best for looking up a concept, API name, or feature by keyword.

輸入參數query必填

get_fonto_page

Get Fonto Page

Fetch the full content of a Fonto documentation page by its slug (the part of the URL after /latest/). Use search_fonto_docs or list_pages first to find the right slug.

輸入參數slug必填

lookup_api

Look Up Fonto API

Look up a Fonto API symbol by exact name (function, class, manager, type, or component — e.g. 'documentsManager', 'createIconWidget', 'CvkOptions') and return its full documentation page in one call. Accepts call syntax ('registerWidget()'), import paths ('fontoxml-families/src/createIconWidget'), and member access ('documentsManager.getNodeById', resolved to the owner's page). Matching is case-sensitive first, so 'DocumentsManager' (class) and 'documentsManager' (instance) resolve to different pages; other same-named pages are listed as alternatives. If no page has that name, returns search suggestions instead — use search_fonto_docs for concepts or partial names.

輸入參數name必填

list_pages

List Fonto Pages

Filter Fonto documentation pages by title, product, or ancestry keyword. Returns all matches without ranking — useful when you know the product area or part of the page title. For full-text relevance search use search_fonto_docs; for the complete catalog use the fonto://catalog resource.

輸入參數keyword必填

版本歷史

5
  1. v1.8.1最新Oct 9, 2026
  2. v1.8.0Oct 8, 2026
  3. v1.7.0Oct 3, 2026
  4. v1.6.4Sep 25, 2026
  5. v1.6.3Sep 16, 2026

評分

已滿足適用檢查項的 100%

該評分衡量此收錄條目的完整程度,依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價,也不表示伺服器安全或目前可用:可連通性是帶日期的檢查結果,評分也不包含安裝量或使用者評價。

  • 近 30 天內可連通必需25 分
  • 已宣告認證方式必需10 分
  • 已列出工具必需15 分
  • 描述必需15 分
  • 圖示10 分
  • 文件15 分
  • 授權條款10 分