Open Knowledge Format
Open Knowledge Format gives teams a plain, portable way to package context, metadata, sources, trust signals, and useful explanations for AI tools.
OKF turns a folder of markdown files into a small knowledge system. People can read it, agents can parse it, and teams can move it between tools without needing a proprietary runtime.
AI systems are often limited by missing or scattered context. OKF gives builders and teams a repeatable way to supply that context before asking an assistant or agent to do important work.
KnowledgeBundles uses OKF as the structure, then adds discovery, commerce, creator accountability, and quality review around the format.
OKF is not another closed knowledge-base product. It is the exchange format underneath the work: a way to make knowledge easier to inspect, version, share, and consume.
Plain-English explanation
Google introduced OKF because useful AI work depends on the context around a task: what a table means, how a metric is calculated, which policy applies, where a claim came from, and whether the information is still current. Their argument is that this context should not be trapped inside one wiki, one catalogue, one app, or one vendor.
The Google explanation says the answer is not another knowledge service. OKF is a shared way to represent knowledge so different producers and consumers can exchange it without a custom integration each time.
An OKF bundle is a directory tree of markdown files. Each concept is one file, and the metadata lives in YAML frontmatter at the top. If a person can open a text file, they can inspect the bundle.
Fields such as type, title, description, resource, and tags help tools route, preview, search, and explain the content instead of treating it as loose text.
The current OKF spec makes provenance, verification, freshness, and lifecycle status first-class. In plain English: the bundle can show what the knowledge came from, who or what checked it, and whether it may be stale.
Most organisations already have useful knowledge, but it is spread across catalogues, wikis, shared drives, code comments, notebooks, and people's heads. Agents then have to reconstruct the same background over and over before they can answer well.
OKF uses formats teams already understand: folders, markdown, frontmatter, links, index files, and logs. That makes the knowledge easier to version, review, search, copy, publish, and load into different workflows.
Google's launch article introduced OKF v0.1. The public spec is now v0.2 and adds more explicit support for provenance, trust, lifecycle, freshness, and attested computation while keeping the basic idea minimal.
A Knowledge Bundle is the commercial, buyer-facing package. OKF is the structure underneath it. The marketplace can then review whether the bundle is readable, source-aware, portable, and useful before sending buyers to download or request one.
Simple shape
The folder gives the bundle structure. The frontmatter gives tools quick metadata. The markdown body explains the concept in human language.
bundle/
index.md
log.md
playbooks/
onboarding.md
references/
source-notes.md
---
type: Playbook
title: Customer onboarding
sources:
- id: policy
resource: /references/source-notes.md
verified:
- { by: human:reviewer, at: 2026-08-16T09:00:00Z }
status: stable
---Google source
Google Cloud introduced Open Knowledge Format as an open, portable way to represent the metadata, context, and curated knowledge that AI systems need. The launch article introduced v0.1; the public spec is now v0.2 and describes a minimal structure based on markdown files, YAML frontmatter, links, provenance, trust, lifecycle, and freshness signals.
Next step