diataxis-mcp-server
Enables agents to classify, audit, and plan documentation using the Diataxis framework with deterministic heuristics and no external LLM calls.
README
diataxis-mcp-server
MVP TypeScript MCP server that helps coding and documentation agents apply the Diataxis framework to documentation work.
The server is deterministic. It uses small heuristics and Zod schemas; it does not call external LLM APIs.
Status
This project is experimental MVP software. The tools are intended to help agents classify, audit, and plan documentation, but the heuristics are conservative and should be reviewed by a human maintainer before large documentation refactors.
Refactor recommendations are advisory. The server suggests extraction targets, navigation groups, and possible file paths, but it does not write, move, rename, crawl, or persist files.
What it provides
Tools:
classify_doc_request: classify a documentation request or draft into tutorial, how-to, reference, or explanation.audit_doc_page: audit one Markdown page for mode fit, mixed-mode signals, extraction targets, and next actions.audit_doc_tree: audit multiple pages and summarize the documentation architecture, missing modes, dominant modes, and suggested navigation. Acceptsscopeasfull-docs,section, orsampleso coverage warnings can be interpreted correctly.suggest_doc_refactor: turn page audits into an advisory, path-aware refactoring sequence.plan_doc_set: propose an initial Diataxis-aligned documentation set from a project summary.
Resources:
diataxis://checklists/tutorialdiataxis://checklists/how-todiataxis://checklists/referencediataxis://checklists/explanationdiataxis://rubrics/classificationdiataxis://workflow/audit-existing-docsdiataxis://workflow/create-docs-from-scratch
Development
npm install
npm test
npm run build
Run the MCP server over stdio:
npm run build
node dist/index.js
Example MCP client command configuration:
{
"mcpServers": {
"diataxis": {
"command": "node",
"args": ["<absolute-path-to-checkout>/dist/index.js"]
}
}
}
IDE configuration examples are included in examples/ and in the npm package for client setup. Run npm run build before using the local dist/index.js configuration.
After this package is published to npm, clients that support npx can use:
{
"mcpServers": {
"diataxis": {
"command": "npx",
"args": ["diataxis-mcp-server"]
}
}
}
Example tool call
Payload for classify_doc_request:
{
"request": "Write a guide that shows how to configure SSO for production.",
"title": "Configure SSO",
"audience": "Operators who already know the product"
}
Shortened example output:
{
"mode": "how-to",
"confidence": 0.58,
"warnings": [],
"rationale": "Detected how-to because the strongest signals match accomplishing a practical task.",
"suggestedStructure": [
"Task goal",
"Prerequisites",
"Procedure",
"Verification",
"Related reference"
]
}
Project layout
src/core: deterministic Diataxis classification, audit, refactor, and planning logic.src/mcp: MCP server, tools, resources, and prompts.src/content: short operational checklists and workflow text.test: focused unit tests and Markdown fixtures.references/diataxis: local source-of-truth notes for Diataxis-derived behavior.
Licensing
This repository uses a split license model.
- Code, build configuration, package metadata, and tests are licensed under Apache-2.0. See
LICENSEfor the full license text andLICENSE-CODE.mdfor the code-license scope note. - Documentation, examples, prompts, checklists, rubrics, workflow text, and Diataxis-derived operational guidance are licensed under CC-BY-SA 4.0. See
LICENSE-CONTENT.md. - The split license model is summarized in
LICENSING.md. - Diataxis attribution notices are recorded in
NOTICE.md.
Attribution
This project contains operational guidance derived from Diataxis.
Diataxis was created by Daniele Procida and is published at https://diataxis.fr/. The upstream repository is https://github.com/evildmp/diataxis-documentation-framework, and the upstream content license is CC-BY-SA 4.0.
This project is not affiliated with or endorsed by the Diataxis project.
推荐服务器
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。
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。
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。
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。
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。
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。
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。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
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。