{
 "acronym": "OKF",
 "adoption": {
  "by": null,
  "level": "unmeasured",
  "probe_run": null
 },
 "aliases": [],
 "analog": "a wiki or knowledge base (Obsidian vault, Notion space) with structured frontmatter, kept in git",
 "collisions": [
  "Open Knowledge Foundation (okfn.org) — an unrelated, pre-existing nonprofit that also uses the initialism OKF"
 ],
 "discovery": {
  "mechanism": "none — a bundle is a directory of markdown files distributed as a git repository, tarball/zip, or subdirectory of a larger repo; there is no fixed web-origin path or well-known convention for locating one",
  "path": "none"
 },
 "dispute": null,
 "evidence": [
  {
   "claim": "Full OKF v0.2 specification: bundle structure, concept documents and frontmatter, provenance/trust/lifecycle fields (`sources`, `generated`, `verified`, `status`, `stale_after`), cross-linking, actor convention, index/log files, Attested Computation concepts, conformance rules (only `type` is required; consumers must not reject unknown fields/types/broken links), and versioning policy",
   "date": "2026-08",
   "label": "VERIFIED",
   "url": "https://raw.githubusercontent.com/GoogleCloudPlatform/open-knowledge-format/main/SPEC.md"
  },
  {
   "claim": "Google Cloud announced OKF v0.1 on 2026-06-12: a directory of markdown files with YAML frontmatter, requiring no SDK or new runtime; shipped with a reference enrichment agent, a static HTML visualizer, and three sample bundles (GA4, Stack Overflow, Bitcoin) explicitly described as 'proofs of concept'",
   "date": "2026-06-12",
   "label": "VERIFIED",
   "url": "https://cloud.google.com/blog/products/data-analytics/how-the-open-knowledge-format-can-improve-data-sharing"
  },
  {
   "claim": "The OKF specification and reference implementation moved out of GoogleCloudPlatform/knowledge-catalog into its own dedicated repository, GoogleCloudPlatform/open-knowledge-format; the old location is marked a frozen, unmaintained snapshot with an explicit notice to stop using it",
   "date": "2026-08",
   "label": "VERIFIED",
   "url": "https://raw.githubusercontent.com/GoogleCloudPlatform/knowledge-catalog/main/okf/README.md"
  },
  {
   "claim": "The dedicated open-knowledge-format GitHub repository was created 2026-08-11 and had 288 stargazers and 11 forks as of this verification",
   "date": "2026-09-05",
   "label": "VERIFIED",
   "url": "https://github.com/GoogleCloudPlatform/open-knowledge-format"
  }
 ],
 "form": "document-format",
 "governance": "Google Cloud open-source project; canonical repository is GoogleCloudPlatform/open-knowledge-format on GitHub (moved out of the GoogleCloudPlatform/knowledge-catalog monorepo in August 2026, which now carries only a frozen, unmaintained snapshot)",
 "group": null,
 "history": [
  {
   "date": "2026-09-05",
   "note": "seeded from Exa Agent research run"
  },
  {
   "date": "2026-09-06",
   "note": "verified against the OKF v0.2 SPEC.md, the original Google Cloud blog announcement, and the repo-migration notice in the old knowledge-catalog/okf directory"
  }
 ],
 "id": "okf",
 "implementation": "prototype",
 "last_verified": "2026-09-06",
 "name": "Open Knowledge Format",
 "native_status": "\"An open specification; contributions, alternative implementations, and adoption beyond Google products are all explicitly welcomed\"",
 "origin": {
  "date": "2026-06-12",
  "org": "Google Cloud (Data Cloud team: Sam McVeety, Amir Hormati)"
 },
 "part_of": null,
 "parties": [
  "agent-site"
 ],
 "phase": "draft",
 "probe": {
  "method": "N/A",
  "on_absent": "No conformant OKF bundle at the given location; this says nothing about whether the org has other, non-OKF internal documentation",
  "parse": "Every non-reserved `.md` file in the tree has a parseable YAML frontmatter block containing a non-empty `type` field (OKF §11 conformance); an optional root `index.md` may declare `okf_version`",
  "path": "none — bundle location is producer-defined (git repo URL, tarball, or subdirectory path); not servable via a fixed {origin} convention"
 },
 "question": "How can an organization package durable, structured knowledge (schemas, metrics, playbooks, runbooks) so both humans and agents can read, produce, and trust it without a bespoke SDK?",
 "scope": "general",
 "spec_url": "https://github.com/GoogleCloudPlatform/open-knowledge-format/blob/main/SPEC.md",
 "step": "read",
 "summary_md": "The Open Knowledge Format (OKF) is a Google-originated specification for representing organizational knowledge — table schemas, metric definitions, playbooks, API references — as a directory of plain markdown files with YAML frontmatter [1][2]. Each unit of knowledge (a \"concept\") is one markdown file; the only required field is `type`, everything else (title, description, tags, and an open-ended set of trust/provenance/lifecycle fields) is optional, and unknown fields or types must be tolerated by any conformant consumer [1]. Concepts cross-link with ordinary markdown links, so a bundle forms a graph, not just a directory tree.\n\nIt solves the problem of fragmented, non-portable \"tribal knowledge\" that today lives in metadata catalogs, wikis, code comments, and senior engineers' heads, and that every organization and every catalog vendor re-solves from scratch with incompatible schemas and SDKs [2]. Because a bundle is just files, it is git-diffable, requires no proprietary API to read or write, and formalizes an already-emerging pattern (the \"LLM-wiki\": Obsidian vaults, `AGENTS.md`/`CLAUDE.md` files, ad hoc `index.md`/`log.md` conventions) rather than inventing a new one [2]. v0.2, the current version, adds first-class provenance (`sources`, with per-source credibility signals like `usage_count` and `last_modified`), trust tiers derived from `verified` events, lifecycle fields (`status`, `stale_after`), and a dedicated \"Attested Computation\" concept type that lets a value's sanctioned computation be checked against what actually ran [1].\n\nWhat it does not do: define a fixed taxonomy of concept types, prescribe any storage/serving/query infrastructure, replace domain-specific schemas such as Avro or OpenAPI (it references them, not subsumes them), or specify a packaging/invocation standard for the code an executor or attester points at [1]. It also has no discovery mechanism of its own — there is no well-known path or fixed convention for finding a bundle at a web origin; a bundle's location is whatever the producer chooses (a git repo URL, a tarball, a subdirectory).\n\nCurrent state, as of this verification (September 2026): OKF v0.1 was announced 2026-06-12 and v0.2 followed by August 2026, when the project also moved out of the `GoogleCloudPlatform/knowledge-catalog` monorepo into its own dedicated repository, `GoogleCloudPlatform/open-knowledge-format`; the old location now carries only a frozen, explicitly unmaintained snapshot [3]. The dedicated repo, created 2026-08-11, had 288 GitHub stars and 11 forks as of 2026-09-05 [4]. All shipped tooling — the reference enrichment agent (BigQuery-to-OKF) and the static HTML graph visualizer — is explicitly described by Google as a \"proof of concept,\" and the only public examples of conformant bundles are the sample datasets Google itself produced (GA4, Stack Overflow, Bitcoin, and a fourth 'Acme Retail' sample added after the initial announcement) [1][3].\n\nWhat is disputed or unknown: no adoption beyond Google's own samples and Google Cloud's Knowledge Catalog integration was found; there is no independent producer or consumer implementation, no third-party adoption count, and no evidence yet of OKF being used to back a production agent system outside Google. \"OKF\" also collides with the pre-existing Open Knowledge Foundation (okfn.org), an unrelated nonprofit; care is needed to distinguish the two in any listing.",
 "track": "single-vendor",
 "version": {
  "date": "2026-08",
  "label": "v0.2"
 },
 "url": "https://agenticweb.wiki/protocols/okf/",
 "category": "read",
 "type": "protocol"
}