Google OKF: The Simple Folder That Gives AI Agents Your Entire Company Data

Google OKF: The Simple Folder That Gives AI Agents Your Entire Company Data

Understanding Active Users and Knowledge Management

The Challenge of Defining Active Users

  • Companies often struggle to define "active users" due to scattered knowledge across various platforms.
  • Information is found in Confluence pages, Notion docs, Slack threads, and even within individual employees' minds.

Context Assembly Issues

  • Every AI agent must repeatedly search for and interpret the same information, leading to inefficiencies.
  • Google refers to this as "solving context assembly from scratch." A unified format could streamline this process.

Introducing Open Knowledge Format (OKF)

What is OKF?

  • The Open Knowledge Format (OKF) is a new specification from Google Cloud that simplifies knowledge management by using plain Markdown files in a Git repository.
  • Each file represents one concept (e.g., a table or metric), with minimal required metadata—only the type field is mandatory.

Benefits of OKF

  • Agents can access business knowledge easily without complex retrieval systems; they read it like a handbook.
  • This approach allows agents to understand company operations without needing extensive background knowledge.

Institutional Memory and Its Importance

The Role of Institutional Memory

  • Modern agents excel at tasks but lack the institutional memory that experienced employees possess, which includes specific facts about metrics and processes.
  • Teams currently rebuild context manually for every project, leading to redundancy and inefficiency in data modeling.

Structure of an OKF Document

Building an OKF Document

  • An OKF document consists of structured YAML metadata followed by free-form Markdown content detailing schemas or formulas relevant to the topic at hand.
  • Metadata helps agents filter and route information effectively while allowing flexibility in additional fields such as descriptions or links.

Navigable Graph of Business Knowledge

Linking Documents Together

  • Documents within an OKF framework link through standard Markdown links, creating a navigable graph representing business knowledge comprehensively.
  • Bundles consist of directories containing related documents organized intuitively according to team needs rather than rigid schemas.

Version Control and Maintenance

Managing Knowledge Like Code

  • The OKF structure allows for version control similar to code repositories, enabling tracking changes over time through pull requests and logs.
  • It accommodates broken links or unknown fields gracefully, reflecting real-world scenarios where knowledge may be incomplete or evolving.

Practical Implementation of OKF

Automating Documentation Creation

  • Google provides tools that automate the creation of OKF documents by analyzing existing datasets and documentation for completeness before human review occurs.

Sample Bundles Available

  • Ready-made bundles are available for common datasets like Google Analytics or Bitcoin data, facilitating quick adoption into workflows.

Key Principles Behind OKF

Core Principles

  • The format emphasizes independence between producers and consumers while being minimally opinionated with only one required field: type.
  • Unlike traditional wikis designed primarily for human readers, OKFs are machine-readable text files that can be utilized anywhere.

(t=406] Conclusion on Adoption Challenges

Current State of Adoption

  • As an early-stage initiative (version 0.1), there are challenges regarding freshness; live data should still be managed via APIs rather than static documents.
  • Emphasizes writing down stable expertise once so all agents can access it efficiently without redundant explanations.
Video description

How do you give an AI agent your company's exact context without building an expensive RAG pipeline or Vector Database? Enter Google OKF (Open Knowledge Format v0.1)—a new open standard that packages your organizational knowledge as a simple, version-controlled directory of markdown files. In this video, Cloud Codes provides a practical guide on how to structure company knowledge for AI agents using OKF. We break down the anatomy of an OKF file (YAML frontmatter + Markdown body), how markdown links create a navigable graph for agents, and how "progressive disclosure" keeps your LLM context window clean. We also explore how Google's BigQuery enrichment agent can auto-draft these files for you, why this "LLM Wiki" approach beats traditional human-only wikis like Notion or Confluence, and exactly how OKF pairs perfectly with MCP (Model Context Protocol). If MCP is the socket, OKF is the data flowing through it. ⏱️ TIMESTAMPS: 0:00 - The AI Agent Context Problem (Data Silos) 0:47 - What is Google OKF (Open Knowledge Format)? 1:07 - Why AI Agents Need Structured Memory 1:48 - Andrej Karpathy's "LLM Wiki" Concept 1:63 - The Anatomy of an OKF File (YAML + Markdown) 2:21 - Building a Knowledge Graph (Links & Bundles) 2:46 - Progressive Disclosure & Log Files 2:87 - Google's BigQuery Enrichment Agent 3:64 - OKF vs. Wikis & RAG (Vector Databases) 3:94 - OKF vs. MCP (Model Context Protocol) 4:20 - The Honest Verdict on v0.1 #googleokf #aiagents #systemdesign #rag #mcp #softwareengineering #googlecloud #cloudcodes 👇 SUBSCRIBE & WATCH NEXT Subscribe for a new systems deep-dive every week: https://www.youtube.com/channel/UC0DZj1PNa_Fp0MU6uPSKv5w?sub_confirmation=1 📱 CONNECT WITH US Twitter/X: x.com/cloud_codes Join our developer community: discord.gg/HVnH9SY48 User Queries: what is google okf open knowledge format tutorial how to structure data for ai agents google okf vs mcp model context protocol vs okf okf vs rag vector database how to build an llm wiki google cloud okf markdown explained give ai agents company context replace confluence with ai wiki