Google Cloud released the Open Knowledge Format (OKF) v0.1, an open-sourced standard for organizing AI-agent knowledge.

Made with Rinto — analyse your own content free

OKF v0.1 Release Details

OKF v0.1 was released on June 12, 2026, and open-sourced under the Apache 2.0 license on GitHub.

Addresses Enterprise Knowledge Fragmentation

OKF focuses on solving fragmented enterprise knowledge, which is hard to reuse and deliver consistently to AI agents.

Core Problems OKF Solves

OKF addresses critical issues in the AI agent stack, particularly knowledge delivery fragmentation.

Basic Definition and Specification

OKF standardizes emerging knowledge representation patterns into a common specification usable across teams and organizations.

OKF Co-Authors

OKF was co-authored by Sam McVeety, Google Cloud Data Cloud Tech Lead, and Amir Hormati, BigQuery Tech Lead.

OKF Core Design Principles

OKF's core design uses plain Markdown, file directories for organization, and YAML frontmatter for metadata.

Easy to Use and Version

This design makes OKF easy to read, edit, store, and version, as any text editor can open it and Git can track its history.

Knowledge Bundle Structure

The basic deployment unit of OKF is a Knowledge Bundle, an independent directory containing multiple Markdown concept files.

YAML Frontmatter Rules

OKF requires only one mandatory metadata field in its YAML frontmatter.

OKF's Complementary Positioning

OKF is designed as a complementary knowledge layer, not a replacement for existing AI infrastructure standards like MCP or RAG.

OKF vs Model Context Protocol (MCP)

MCP is a dynamic communication protocol, while OKF stores static, curated, persistent knowledge.

Standard	Role
MCP	Dynamic tool communication
OKF	Stable knowledge content

OKF vs Retrieval-Augmented Generation (RAG)

RAG focuses on semantic retrieval from raw documents, whereas OKF targets manually reviewed, structured knowledge bundles.

Approach	Focus
RAG	Semantic retrieval from large volumes of raw documents
OKF	Manually reviewed, structured, authoritative bundles

OKF vs Repository Conventions

Repository-local conventions like AGENTS.md describe AI agent behavior within a codebase; OKF generalizes this cross-system.

Official Reference Implementations

Google Cloud released three reference implementations with OKF v0.1 to demonstrate the full workflow from generation to reuse.

Knowledge Enrichment Agent

This automated tool generates OKF knowledge files from BigQuery assets and enriches them using LLMs.

Static HTML Visualization Tool

This second reference tool is a static HTML visualization tool that renders interactive knowledge graphs.

Open-Source Sample Bundles

Google Cloud released three sample OKF bundles on GitHub, including Google Analytics 4 datasets and Public Stack Overflow data.

Three Core Design Principles

OKF follows three consistent design principles: minimal constraints, separation of producer and consumer, and format-first architecture.

Industry Trend Reflected by OKF

The release of OKF signifies a deeper shift in the AI agent ecosystem, emphasizing reliable access to authoritative business context.

Conclusion

Google Cloud's OKF v0.1 is a lightweight but important release, formalizing community-proven patterns into a vendor-neutral format.

▸ 9 Expand

APEX

Google Cloud Launches Open Knowledge Format

Google Cloud released the Open Knowledge Format (OKF) v0.1, an open-sourced standard for organizing AI-agent knowledge.

Made with Rinto — analyse your own content free

EVID

OKF v0.1 Release Details

OKF v0.1 was released on June 12, 2026, and open-sourced under the Apache 2.0 license on GitHub.

▸ 1 Expand

SUP

Addresses Enterprise Knowledge Fragmentation

OKF focuses on solving fragmented enterprise knowledge, which is hard to reuse and deliver consistently to AI agents.

JUST

Fragmented Knowledge Problem

Internal knowledge is scattered across various sources, leading to custom, non-reusable context delivery pipelines for each team.

▸ 4 Expand

SECT

Core Problems OKF Solves

OKF addresses critical issues in the AI agent stack, particularly knowledge delivery fragmentation.

SUP

Knowledge Not Agent-Readable

Most enterprise knowledge is not stored in one clean, agent-readable system, existing in disparate formats like wikis or spreadsheets.

▸ 2 Expand

RISK

Fragmentation Bottleneck for Agents

Fragmented knowledge becomes a bottleneck when an agent needs to answer business questions, such as calculating user counts.

EXMP

Example: User Count Calculation

An agent may struggle to determine 'How is weekly active user count calculated?' due to knowledge fragmentation.

EXMP

Example: Database Order Status

An agent might need to know 'Which tables define order status?' but finds the information scattered.

RISK

Increased Hallucination Risk

Without standardized knowledge input, agents receive incomplete context, increasing hallucination risk and requiring custom parsers.

EVID

Inspired by LLM Wiki Concept

Andrej Karpathy's LLM Wiki concept, where LLMs maintain structured knowledge, provides important background for OKF.

▸ 5 Expand

SECT

Basic Definition and Specification

OKF standardizes emerging knowledge representation patterns into a common specification usable across teams and organizations.

SUP

OKF Co-Authors

OKF was co-authored by Sam McVeety, Google Cloud Data Cloud Tech Lead, and Amir Hormati, BigQuery Tech Lead.

▸ 3 Expand

FRMW

OKF Core Design Principles

OKF's core design uses plain Markdown, file directories for organization, and YAML frontmatter for metadata.

SUP

Uses Plain Markdown

OKF uses plain Markdown for its content.

SUP

Organizes as File Directories

OKF organizes knowledge as file directories.

SUP

Uses YAML Frontmatter

OKF uses lightweight YAML frontmatter for structured metadata.

INSG

Easy to Use and Version

This design makes OKF easy to read, edit, store, and version, as any text editor can open it and Git can track its history.

▸ 2 Expand

SUP

Knowledge Bundle Structure

The basic deployment unit of OKF is a Knowledge Bundle, an independent directory containing multiple Markdown concept files.

EVID

Concept File Representation

Each .md file within a bundle represents one business concept, such as a BigQuery table or an API endpoint.

EVID

File Path as Concept ID

Within a bundle, the file path acts as the unique ID of the concept, making the knowledge structure portable.

▸ 4 Expand

SUP

YAML Frontmatter Rules

OKF requires only one mandatory metadata field in its YAML frontmatter.

REQ

Mandatory 'type' Field

The mandatory 'type' field identifies the category of the concept, for example, a database table or a metric.

EVID

Recommended Optional Fields

Optional recommended metadata fields include title, description, resource, tags, and timestamp.

JUST

Minimal Requirement Intentional

This minimal requirement provides interoperability structure without forcing organizations to adopt a fixed internal document model.

SUP

Cross-Concept Relationships

Cross-concept relationships are represented through standard Markdown links, allowing navigable knowledge graphs without proprietary indexing.

▸ 3 Expand

SECT

OKF's Complementary Positioning

OKF is designed as a complementary knowledge layer, not a replacement for existing AI infrastructure standards like MCP or RAG.

CMPR

OKF vs Model Context Protocol (MCP)

MCP is a dynamic communication protocol, while OKF stores static, curated, persistent knowledge.

Standard	Role
MCP	Dynamic tool communication
OKF	Stable knowledge content

▸ 1 Expand

CMPR

OKF vs Retrieval-Augmented Generation (RAG)

RAG focuses on semantic retrieval from raw documents, whereas OKF targets manually reviewed, structured knowledge bundles.

Approach	Focus
RAG	Semantic retrieval from large volumes of raw documents
OKF	Manually reviewed, structured, authoritative bundles

RISK

RAG Pipeline Limitations

RAG pipelines may include noisy documents, duplicate content, inconsistent terminology, and outdated information.

▸ 1 Expand

SUP

OKF vs Repository Conventions

Repository-local conventions like AGENTS.md describe AI agent behavior within a codebase; OKF generalizes this cross-system.

INSG

Reusable Shared Enterprise Knowledge

A single OKF bundle can be used across many repositories, agents, dashboards, and data tools, making it suitable for shared enterprise knowledge.

▸ 3 Expand

SECT

Official Reference Implementations

Google Cloud released three reference implementations with OKF v0.1 to demonstrate the full workflow from generation to reuse.

▸ 3 Expand

COMP

Knowledge Enrichment Agent

This automated tool generates OKF knowledge files from BigQuery assets and enriches them using LLMs.

STEP

Draft Markdown Creation

The agent scans tables and views, then creates draft Markdown concept files for each data asset.

STEP

LLM Enrichment Process

A secondary LLM step enriches files by reading authoritative documents and adding details like table descriptions or business context.

INSG

Reduces Manual Documentation

This process reduces manual documentation work, converting existing data assets into agent-readable knowledge bundles.

▸ 2 Expand

COMP

Static HTML Visualization Tool

This second reference tool is a static HTML visualization tool that renders interactive knowledge graphs.

JUST

Local Processing for Privacy

The tool has no backend dependency, with all parsing and rendering happening locally in the browser, ensuring enterprise privacy.

INSG

Compliance and Data Sensitivity

Knowledge data is not uploaded to external servers, allowing local inspection of sensitive business data and private documentation.

▸ 1 Expand

COMP

Open-Source Sample Bundles

Google Cloud released three sample OKF bundles on GitHub, including Google Analytics 4 datasets and Public Stack Overflow data.

INSG

Aids Developer Understanding

These samples help developers understand OKF concept files, metadata conventions, Markdown linking patterns, and knowledge graph structures.

▸ 3 Expand

SECT

Three Core Design Principles

OKF follows three consistent design principles: minimal constraints, separation of producer and consumer, and format-first architecture.

▸ 1 Expand

SUP

Minimally Opinionated Standard

OKF makes only the 'type' field mandatory, leaving everything else flexible for teams to define their own metadata and logic.

JUST

Ease of Adoption

This approach makes OKF easier to adopt, requiring minimal structural additions for cross-system recognition, not a full content model rebuild.

▸ 1 Expand

SUP

Separation of Producer and Consumer

OKF separates knowledge creation from knowledge consumption, enabling a bidirectional workflow where humans write and machines generate.

INSG

Independent Evolution

Knowledge producers and consuming agents can change independently while the file format remains stable.

▸ 1 Expand

SUP

Format Rather Than Platform

OKF is a file specification, not a hosted platform, requiring no proprietary SDKs, paid runtimes, or specific LLM frameworks.

INSG

Reduces Vendor Lock-In Risk

This design greatly reduces vendor lock-in risk, as existing OKF bundles remain readable and portable even if vendor support ceases.

▸ 4 Expand

SECT

Industry Trend Reflected by OKF

The release of OKF signifies a deeper shift in the AI agent ecosystem, emphasizing reliable access to authoritative business context.

INSG

New Bottleneck: Reliable Context

The main bottleneck is no longer solely base model reasoning; it's reliable access to authoritative business context for enterprise use cases.

RISK

Poor Results from Incomplete Context

Even advanced models can produce poor results when internal context is scattered, outdated, or incomplete.

INSG

Reusable Infrastructure

OKF transforms ad-hoc knowledge engineering practices, such as manual cleanup and team-specific conventions, into reusable infrastructure.

INSG

Middleware Layer in AI Architecture

From a full-stack AI architecture perspective, OKF can function as a middleware layer between enterprise data systems and AI agents.

▸ 3 Expand

SECT

Conclusion

Google Cloud's OKF v0.1 is a lightweight but important release, formalizing community-proven patterns into a vendor-neutral format.

INSG

Shift in AI Agent Competition

AI agent competition is moving beyond model performance alone, with standardized knowledge infrastructure becoming key to production-grade systems.

INSG

Benefits for Enterprises

For enterprises, OKF can reduce knowledge silos, improve context quality, and lower repeated engineering costs in multi-agent architectures.

INSG

Future Common Format

As more data platforms and agent frameworks adopt OKF, it may become a common format for structured knowledge circulation in the AI ecosystem.