Loading source
Pulling the file list, source metadata, and syntax-aware rendering for this listing.
Source from repo
Optimizes content to be cited and surfaced by AI search engines like ChatGPT, Perplexity, Google AI Overviews, and Claude.
Files
Skill
Size
Entrypoint
Format
Open file
Syntax-highlighted preview of this file as included in the skill package.
references/okf.md
1# Open Knowledge Format (OKF)23Google's v0.1 markdown spec for representing site content as an agent-readable bundle. Introduced on the [Google Cloud blog](https://cloud.google.com/blog/products/data-analytics/how-the-open-knowledge-format-can-improve-data-sharing) on 2026-06-12 and shipped inside Knowledge Catalog.45## What it is67OKF is a directory of cross-linked markdown files. Each file has:89- A YAML frontmatter block (`type` required; `title`, `description`, `resource`, `tags`, `timestamp` recommended)10- A standard markdown body11- Standard markdown links to other files in the bundle (which the spec treats as concept relationships)1213An optional `index.md` lists the files for progressive disclosure. The bundle can be distributed as a git repo (recommended), a tarball/zip, or a subdirectory of a larger repo.1415The [full spec](https://github.com/GoogleCloudPlatform/knowledge-catalog/blob/HEAD/okf/SPEC.md) fits on one page. The repo lives under `GoogleCloudPlatform` (the "not an official Google product" disclaimer is Google's standard open-source boilerplate, not a denial — it appears on most of Google's open-source repos including their main AI samples repo).1617### A minimal concept file1819```markdown20---21type: Article22title: How to Connect the Ahrefs MCP Server to Manus23description: The official MCP servers, why they did not connect, and the fix.24resource: https://yoursite.com/blog/ahrefs-mcp-manus/25tags: [mcp, ahrefs]26---2728# How to Connect the Ahrefs MCP Server to Manus2930The body of the post, as clean markdown.31```3233Add an `index.md` that lists all files so an agent can see the bundle's shape before opening each file, and that is the entire format.3435## Honest framing3637**Google built OKF for data teams sharing catalog metadata** — BigQuery tables, API endpoints, metrics, playbooks. Most of the spec's examples are data-team artifacts, not blog posts. Google's blog post framing: "improve data sharing" and "standardized documentation" for collaboration across teams.3839Pointing OKF at a marketing site is a **clever repurposing** popularized by [Suganthan Mohanadasan](https://suganthan.com/blog/open-knowledge-format/). It's a legitimate use case for the format but not Google's primary one. Frame it accurately when explaining it to founders or marketing teams.4041## What it does for AI search today4243Nothing immediate. Nothing crawls the web for OKF bundles yet — the spec is weeks old, no AI engine has announced integration, and Knowledge Catalog ingests bundles only for paying enterprise customers' data teams.4445Treat OKF as **protocol-layer registration** — the same shape of bet as early `schema.org` adoption was a decade ago. Schema took the better part of ten years to pay off; people who shipped it early are still glad they did.4647A secondary benefit that pays off today regardless: **generating the bundle is itself an internal-linking audit**. Suganthan's tool draws every page as a node and every internal link as an edge, so islands and orphans become obvious at a glance.4849## Where OKF fits in the agent-readable stack5051| Layer | Purpose |52|---|---|53| `sitemap.xml` | Tells a crawler which URLs exist |54| `robots.txt` (with AI bot rules) | Permits or blocks AI crawlers |55| `llms.txt` | Points an agent at the handful of pages you most want read |56| `/pricing.md` | Structured pricing for agent-buyer comparisons |57| **`/okf/` bundle** | Hands over the content itself as cross-linked concepts |58| Schema markup | Per-page structured data (Article, FAQPage, Product, etc.) |5960These stack rather than compete. `llms.txt` is a signpost, OKF is the library.6162## How to ship one6364Three options, ordered by how much effort they take:6566### 1. Suganthan's free web tool (recommended for most sites)6768[suganthan.com/okf-generator](https://suganthan.com/okf-generator/) — paste a URL or sitemap, crawls up to 100 pages, returns a downloadable bundle. Also draws the resulting page graph so you can spot disconnected pages before publishing.6970### 2. WordPress plugin (pending wp.org approval)7172Suganthan's plugin (free, GPL, awaiting wp.org approval at time of writing) installs in a minute, serves the bundle at `/okf/`, and rebuilds on every publish or edit so it stays in sync. Direct download link is in [his blog post](https://suganthan.com/blog/open-knowledge-format/). Requires WordPress 6.0+ and PHP 7.4+. Read-only — never edits posts or settings.7374### 3. By hand7576Only practical for a handful of pages. Each post becomes a markdown file with frontmatter that you cross-link manually. Miserable for a whole site.7778## Hosting & discovery7980Serve the bundle at `yoursite.com/okf/`, starting with `yoursite.com/okf/index.md`:8182- **Static hosts / Cloudflare**: drag and drop83- **WordPress**: Suganthan's plugin handles the serving84- **Static sites with custom paths**: upload the directory to `/okf/`85- **Closed platforms (Wix, Squarespace, most page-builders)**: you usually can't serve files at custom paths — skip OKF entirely8687After it's serving, add a line to `llms.txt` pointing to the bundle so agents that read `llms.txt` (today) can discover the bundle (later).8889## When to skip9091- Site is <10 pages — overhead exceeds payoff92- Site is on a closed platform that won't allow custom paths93- You're not maintaining `llms.txt`, schema markup, or other machine-readable files (OKF compounds with those; alone it does nothing)94- You can't budget the 30 minutes a quarter to refresh the bundle as content changes9596## What to watch9798OKF is v0.1, weeks old. Worth tracking, not worth obsessing over:99100- Whether Google announces OKF support in AI Overviews / Knowledge Graph (currently no signal)101- Whether non-Google engines (ChatGPT, Perplexity, Claude) announce OKF reading102- Whether the spec moves to v1.0 (breaking changes are possible at <1.0)103- Whether Knowledge Catalog adds public ingestion endpoints104- Adoption signals — search GitHub for `okf/index.md` to see who's shipping bundles105