Meet Blume: An Open-Source, Zero-Config Documentation Framework That Ships AI-Ready Docs From a Markdown Folder 认识 Blume：一个开源、零配置的文档框架，从 Markdown 文件夹生成 AI 就绪的文档

TL;DR

• Blume is an open-source, zero-config documentation framework by OpenAI expert Hayden Bleasel that converts Markdown/MDX folders into production-ready sites using a hidden Astro engine.
• It features native "AI-Ready" capabilities including llms.txt generation, per-page raw Markdown access, and a built-in Model Context Protocol (MCP) server for agent integration.
• The tool offers a seamless upgrade path via blume eject, allowing developers to transition from a managed CLI experience to a standalone Astro application while retaining core dependencies.
• Performance is optimized for Core Web Vitals by shipping no client-side JavaScript in the core theme, relying instead on static HTML rendering.
• It supports diverse content sources (filesystem, Notion, remote MDX), internationalization for 36 locales, and integrates with AI SDKs for in-page chat assistants.

Why It Matters

This release signals a significant shift toward documentation systems that are natively designed for both human consumption and machine readability, addressing the growing need for AI agents to reliably access technical resources. By embedding MCP support and standard AI discovery files (llms.txt) directly into the framework, Blume reduces the friction for integrating documentation into agentic workflows. For practitioners, it offers a low-friction entry point for high-performance docs without the maintenance overhead of custom React-based stacks like Docusaurus.

Technical Details

• Architecture: Blume operates as a CLI wrapper around a hidden Astro project. The CLI scans content into a graph, generates an Astro app in .blume/, and uses a single catch-all route to render pages via shipped components.
• AI Integration: Includes a built-in MCP server exposing four read-only tools (search_docs, get_page, list_pages, get_navigation) for IDEs like Cursor and VS Code. It also supports llms.txt flags and in-page AI assistants via the AI SDK.
• Performance: The core theme is static-first, shipping zero client-side JavaScript by default, which ensures strong Core Web Vitals scores. Hot reloading is maintained by rewriting only changed files during development.
• Content Management: Supports multiple content sources including local filesystem, Notion databases, and remote MDX. It handles internationalization with locale-aware routing and RTL layout support.
• Extensibility: Developers can use blume eject to promote the hidden runtime into a standalone Astro app, gaining full control over the underlying infrastructure while keeping the blume package dependency.

Industry Insight

Documentation frameworks are evolving from passive information repositories into active nodes in the AI agent ecosystem. The inclusion of MCP servers and standardized AI discovery protocols suggests that future documentation standards will prioritize machine-readable structures alongside human-friendly UIs. Teams should evaluate their documentation strategies not just for SEO, but for "agent-readability," ensuring their technical assets can be easily ingested by LLMs and coding assistants. Additionally, the trend toward "zero-config" static site generators highlights a demand for reduced operational complexity in developer tooling, favoring solutions that abstract away build pipeline maintenance.

TL;DR

• Blume 是由前 OpenAI 开发者 Hayden Bleasel 发布的开源零配置文档框架，支持将 Markdown/MDX 文件夹直接转换为生产级文档站点。
• 核心架构基于隐藏的 Astro 项目，通过 CLI 自动生成并管理，提供静态优先输出以优化 Core Web Vitals 性能。
• 专为 AI Agent 设计，内置 llms.txt 支持、页面级 Markdown 导出及 Model Context Protocol (MCP) 服务器，便于 Claude Code 等工具直接读取。
• 提供丰富的集成能力，包括 Notion/Sanity 数据源、OpenAPI 规范渲染、多语言支持（36种）及一键“eject”至独立 Astro 应用。

为什么值得看

Blume 解决了传统文档框架配置繁琐的痛点，通过“零配置”理念极大降低了技术文档的搭建门槛。其原生集成 AI Agent 协议（如 MCP 和 llms.txt）标志着文档基础设施正从单纯服务人类读者向同时服务 AI 代理转变，是构建 AI-Native 开发者体验的重要参考。

技术解析

• 架构与引擎：Blume 是一个 CLI 工具加组件库，底层驱动一个隐藏的 Astro 项目。CLI 扫描内容生成图结构，Astro 通过单一 catch-all 路由渲染页面，核心主题不发送客户端框架 JavaScript，确保默认高性能。
• AI 就绪特性：支持通过 llms.txt 和 llms-full.txt 标志为 AI 提供结构化数据；每个页面可通过追加 .md 获取原始 Markdown；内置可选的“Ask AI”助手，兼容 Vercel AI Gateway 等多种后端。
• MCP 服务器集成：可托管 MCP 服务器，暴露 search_docs、get_page、list_pages 和 get_navigation 四个只读工具，使 Claude Code、Cursor 等 IDE 能直接查询文档。
• 工作流与命令：提供完整的生命周期管理，包括 blume init（初始化）、blume dev（热重载开发）、blume build（构建静态站）、blume eject（导出为独立 Astro 应用）以及 blume validate（链接检查）。
• 内容源与扩展：支持文件系统、Notion、GitHub Releases 等多种数据源混合使用；内置对 OpenAPI/AsyncAPI 规范的渲染支持，可生成交互式 API 参考页。

行业启示

• 文档即 API 接口：随着 AI Agent 成为软件开发的一部分，文档不再仅是给人看的，更是给机器读的。内置 MCP 和 llms.txt 支持表明，未来的文档框架必须原生考虑机器可读性。
• 零配置与静态优先的趋势：Blume 通过隐藏复杂性并提供静态优先的输出，反映了开发者对快速启动和高性能交付的追求，这种“开箱即用”且易于退出的设计模式将在更多基础设施中普及。
• 混合内容源的标准化：支持将本地文件与 Notion、Sanity 等远程源统一渲染，表明企业级文档管理正趋向于解耦内容创作平台与前端展示层，通过统一组件库实现跨平台一致性。