---
type: "GlossaryTerm"
title: "Open Knowledge Format (OKF)"
description: "OKF (Open Knowledge Format) is an open, vendor-neutral format that stores organizational knowledge as Markdown files with YAML frontmatter so AI agents can read"
resource: "https://www.contextstudios.ai/glossary/open-knowledge-format-okf"
language: "en"
tags: ["infrastructure"]
generated:
  by: "process:contextstudios-md/1"
  at: "2026-10-08T20:45:53.586Z"
status: "stable"
---

# Open Knowledge Format (OKF)

OKF (Open Knowledge Format) is an open, vendor-neutral format that stores organizational knowledge as Markdown files with YAML frontmatter so AI agents can read curated context directly.

Google Cloud published version 0.1 in June 2026, formalizing the "LLM-wiki" pattern that agent teams had already used informally. An OKF bundle is a directory of linked Markdown files. Each file carries metadata such as type, owner, trust level, and lifecycle status in its frontmatter, and cross-links tie the files into a traversable graph. No SDK, no proprietary runtime, no conversion layer: people and agents read the same file, which is why bundles move between producers and consumers without translation.

Concretely: a data team that maintains 40 metric definitions and 25 join paths in an OKF wiki lets an agent answer the question "which revenue figure is authoritative?" by walking one exact path through the graph, instead of scoring three of five lookalike chunks from a RAG index. That is the distinction to RAG: RAG chops large, unstructured corpora into chunks and retrieves similar passages at query time, while OKF curates stable knowledge — table schemas, metric definitions, runbooks, join paths — in advance so relationships stay intact and auditable. And OKF is not a replacement for MCP either: MCP connects an agent to live tools, while OKF describes what an agent already knows before it starts working; an MCP server can even serve an OKF bundle.