--- type: "WebPage" title: "Open Knowledge Format Generator" description: "Free Open Knowledge Format (OKF) generator. Crawl your site, convert pages to clean markdown, and download a hostable OKF bundle that AI agents can read." resource: "https://toddmorourke.com/tools/okf-generator/" status: "stable" generated: by: "okf-publisher/0.3" at: "2026-07-31T16:39:02+00:00" verified: - by: "human:todd-orourke" at: "2026-07-31T16:39:02+00:00" stale_after: "2027-07-31" timestamp: 2026-07-31T16:39:02+00:00 --- # Open Knowledge Format Generator Updated The OKF Generator below has been updated per Google's v0.2 Spec guidelines. Enter your domain and the generator crawls your sitemap, converts each page to clean markdown, and packages a ready-to-host **Open Knowledge Format (OKF)** bundle you can download as a ZIP. It's the fast way to hand AI agents a structured, readable copy of your site instead of leaving them to scrape your HTML. Domain (the site to convert — e.g. example.com) Generate ▶ See an example — the interactive `viz.html` knowledge graph Every bundle ships a self-contained `viz.html` like this — the one below is toddmorourke.com’s own bundle. Drag a node, scroll to zoom, drag the background to pan, click a node to open the page. No libraries, no third-party requests. ## What is Open Knowledge Format? OKF is a vendor-neutral standard for publishing your content as plain markdown files with YAML frontmatter, organized so AI agents can read and cite it without scraping your HTML. Each file is one concept, a page or a service or an article, with a few structured fields at the top and normal markdown below. The files link to each other, so a bundle is a graph of your knowledge, not just a folder of pages. I run one on this site at [/okf/](https://toddmorourke.com/okf/), and I covered what it is and how to add it to any CMS in [Open Knowledge Format: how to add it to WordPress, Shopify, Webflow, and Squarespace](https://toddmorourke.com/learn/open-knowledge-format/). It exists to make your site legible to agents. When an agent reads a normal page it strips the layout and loses the structure: which page is canonical, how your pages relate, what each thing actually is. An OKF bundle hands that structure over directly. It's the same kind of bet schema markup was a decade ago, machine-readability that compounds rather than spikes. ## What the generator does - **Crawls your sitemap** – finds your XML sitemap and fetches each published page, up to 100. - **Converts to clean markdown** – pulls the main content out of every page and drops navigation, scripts, and layout chrome. - **Builds the full bundle** – an `index.md`, one concept file per page, an `llms.txt` discovery map, a runnable conformance check, and a `README` with hosting steps. - **Validates before download** – checks frontmatter and internal links against the OKF v0.2 spec, reports how much of your bundle carries provenance and how much is still unverified, then packages everything as a ZIP. - **Optional business-entity layer** – with your own Anthropic API key, extracts a structured profile (services, locations, FAQs) from the crawled pages. - **Knowledge graph** – a self-contained `viz.html` that shows how your pages link together. Everything runs in your browser. The bundle is yours to host at `/okf/` on your own domain, and the included `README` walks through the one line to add so agents can find it. ### About the trust signals OKF v0.2 added provenance to the format, and the interesting part is that absence now means something: a concept with no trust signals reads as unverified rather than neutral. Every file this generator writes records what produced it, when the content last changed, and what it was derived from. None of them claims `verified`, because crawling a page isn't reviewing it and no automated tool can honestly say a human checked your content. The `README` gives you the one block to add once you have, which moves a concept to the top trust tier. That line is the only thing in the bundle a tool can't write for you, and I break the whole model down in [what OKF v0.2 changed](https://toddmorourke.com/learn/open-knowledge-format/#okf-v02-trust-signals). The bundle also ships an *attested computation*: a small script you can run with Node that re-counts the concepts and trust tiers straight from the files. It means the numbers the tool reports are reproducible rather than something you have to take on faith. ### About the business-entity layer This step is optional and uses your own Anthropic API key, which goes straight from your browser to Anthropic and never touches my server. Claude reads the crawled pages and extracts a structured profile, but only facts that actually appear in your content. It won't invent a phone number or a service you don't offer, which is the difference between a business profile you can publish and a confident guess. **Important:** OKF is at v0.2 and still early. Google has updated its Cloud Knowledge Catalog to ingest OKF and serve it to their agents, but no answer engine is known to crawl a public `/okf/` folder yet, so this is about being early rather than instant traffic. Extraction is good, not perfect, and quality depends on how cleanly your site is built. JavaScript-only sites and pages behind a login can't be read. The cap is 100 pages, which is enough for most sites. If yours is bigger, run it against a subdomain that holds the section you care about (your blog or docs), so the 100 it bundles are the ones you care about. Want the strategy behind this, not just the files? See my [AEO content strategy](https://toddmorourke.com/learn/aeo-content-strategy/) for how agent-readable content earns citations, the sibling [ARD Checker](https://toddmorourke.com/tools/ard-checker/) for whether agents can find your site at all, and [how I built an AI second brain](https://toddmorourke.com/learn/ai-second-brain/) on the same format.