Getting Started

Run OpenNav after your existing static build command. The output folder must already contain real prerendered HTML such as index.html, docs/getting-started/index.html, or docs/api/index.html.

Install

Install OpenNav in the project that builds your static site.

npm

npm install @opennav-ai/opennav

pnpm

pnpm add @opennav-ai/opennav

yarn

yarn add @opennav-ai/opennav

bun

bun add @opennav-ai/opennav

CLI Quick Start

opennav build --static --output dist --site-url https://example.com --site-name "Example Docs"

Use --dry-run first when you want to preview the exact files OpenNav would create, modify, or skip.

opennav build --static \
  --output dist \
  --site-url https://example.com \
  --site-name "Example Docs" \
  --dry-run

Where It Runs

OpenNav is designed for static site platforms first. If your deployment is a folder of files, OpenNav can run before you upload that folder.

Platform style	Example output	How OpenNav fits
Cloudflare Pages	dist/ or framework output	Run OpenNav after the build with --platform cloudflare-pages, then deploy the same folder.
Netlify	dist/, build/, or framework output	Add OpenNav as the final build step before publish.
Vercel static output	out/ or generated assets	Run OpenNav for exported static routes before upload.
GitHub Pages or CDN hosting	Plain HTML folder	Run OpenNav locally or in CI, then publish the generated files.

Framework SDKs

Use the TypeScript SDK when OpenNav should run inside a framework build or a custom Node script. Static Astro and Next.js export builds are the first framework-specific SDK paths.

Framework	Current support	Output folder
Astro	Static build helper.	dist/
Next.js	Static export helper for output: "export".	out/

More framework helpers will follow in coming releases.

Server-Side Frameworks

OpenNav supports runtime HTML-to-Markdown content negotiation for any Node.js or WinterCG-compatible server. Import OpenNavServer from @opennav-ai/opennav/server and wrap your existing route handlers:

import { OpenNavServer } from "@opennav-ai/opennav/server";

const opennav = new OpenNavServer();

app.get("/docs/:slug", async (c) => {
  const htmlResponse = await renderPage(c.req.param("slug"));
  const result = await opennav.negotiate({
    request: c.req.raw,
    htmlResponse,
  });
  if (result.isErr()) return c.text("Internal error", 500);
  return result.value;
});

When the client sends Accept: text/markdown, the HTML is converted to Markdown on-the-fly. Browser requests still receive normal HTML, and responses include Vary: Accept so caches keep both representations separate.

See the Server-Side Frameworks guide for framework-specific advice.