Docora brings Docus-style documentation to Next.js and React, with MCP server and llms.txt for AI tools

What Docora is

Docora is a new open-source documentation framework aimed at Next.js and React projects. According to the author's announcement on dev.to, it takes its inspiration from Docus, the documentation toolkit in the Nuxt ecosystem, but is written for the React stack from scratch. The code is published on GitHub under the StaticMania organisation, and a live demo is deployed on Vercel.

The pitch is a batteries-included docs site: put content in, get a production-ready site out, with the pieces that normally require extra services — search, SEO plumbing, internationalisation — handled by the framework itself.

Core features

According to the dev.to post, the feature set covers:

• File-based routing: every file placed under the content/ directory becomes a page route automatically.
• MDC syntax: authors write Markdown and get richer components — callouts, cards, tabs and step lists — without adding imports.
• Built-in search: a command palette generated at compile time, with no third-party hosted search service in the loop.
• SEO defaults: sitemap, robots rules, canonical links and auto-generated Open Graph images.
• Internationalisation: one folder per locale, with hreflang tags emitted for search engines.
• AI-facing endpoints: every deployed site exposes an MCP server at /mcp plus an auto-generated llms.txt file, so tools such as Cursor and Claude can read the documentation directly instead of scraping pages.

There is also an optional Ask AI chat widget, built with the AI SDK and routed through Vercel's AI Gateway.

Getting started

Onboarding is a single command — npx create-docora my-docs — followed by npm run dev inside the new project. Beyond that, the author is openly asking the community for feedback on developer experience, missing features and bugs, which signals the project is still early and shaped by its users.

The catch with Ask AI

The one pain point disclosed in the announcement is infrastructure. Docora is built entirely on Vercel's stack — Next.js, deployment on Vercel, and AI Gateway for the chat feature — and the free AI Gateway tier currently caps usage. As a result, the author reports that most visitors to the live demo cannot fully try Ask AI. The post asks readers for pointers on Vercel's Open Source Program or similar support, so the chat feature can stay openly usable for the Next.js community to experiment with and learn from rather than being throttled for most visitors.

Why it matters

Documentation tooling has been an awkward gap in the React world. The strongest dedicated options have historically lived elsewhere — Docus in the Nuxt ecosystem, VitePress in Vue land, Docusaurus on its own React-based toolchain — while Next.js projects often end up assembling a docs setup from raw MDX and a pile of plugins. Docora is an attempt to close that gap with the same convention-over-configuration ergonomics those tools already offer, and its quickstart suggests the barrier to trying it is close to zero.

The more novel part is the AI-native angle. Shipping an MCP server and llms.txt by default treats AI coding assistants as first-class consumers of documentation, alongside browsers and search crawlers. If that pattern spreads, "is your docs site machine-readable" becomes as standard a question as "do you have a sitemap".

The project is young, and its long-term viability depends on community uptake and, for Ask AI specifically, on finding a sustainable hosting arrangement. But as a signal of where docs tooling is heading — compile-time everything, content written as much for agents as for humans — it is worth watching.