Google Cloud’s Open Knowledge Format Turns Docs Into Fuel for AI Agents
Google Cloud has made a new tool. It is called the Open Knowledge Format. It is a simple way to save what a company knows. The goal is to let AI agents read and use that knowledge. An AI agent is a software helper that can do tasks by itself. For example, it can answer hard questions or pull facts from a database (a place where data is stored). The new tool is often called OKF for short. It takes messy notes and files and turns them into clean files. Both machines and people can read these files. Google Cloud shared the news on June 14, 2026. This first version is called OKF v0.1.
The idea is small but strong. Today, knowledge gets locked inside many different tools. With OKF, you write it as plain text files instead. Then any agent, on any cloud, can pick it up and learn from it.
The problem: knowledge is scattered everywhere
Most companies do not keep their know-how in one place. Useful facts are spread out. Some sit in data catalogs. A data catalog is like a big list that names a company’s data sets. Other facts hide in wikis, in code notes, in notebooks, or just inside one expert’s head.
This mess is a real problem for AI agents. When an agent needs facts, it must dig through many systems. Here is an example. Say an agent must write an SQL query for some data. SQL is the language used to ask questions of a database. To do this well, the agent needs to know what the data means. It also needs to know how the data is set up. If that knowledge is messy and spread out, the agent slows down. It may even guess wrong.
Google Cloud built OKF to fix this exact gap. The goal is simple. It brings scattered knowledge into one tidy, shared shape.
How the Open Knowledge Format works
OKF is built on a format you may know: Markdown. Markdown is a plain-text way of writing. It uses simple marks for headings, links, and lists. It is easy to read. It also works almost anywhere.
An OKF bundle is just a folder of Markdown files. Each file has a small block at the top. This block is called YAML frontmatter. Think of it as a quick label. It holds facts about the file. Only one field is required. That field is “type.” It says what kind of knowledge the file holds. Everything else is optional. That can include a title, a short description, a link, tags, and timestamps. Timestamps are the dates that show when a file was made or changed.
The files link to each other using normal Markdown links. When many files link together, they form a knowledge graph. A knowledge graph is a web of facts. It shows how the pieces of information relate. This lets an agent follow a trail from one idea to the next.
The files are plain text, so they are friendly to everyone. You can open them in any editor. They look nice on GitHub. GitHub is a popular site where developers share code. Search tools can also list them with no trouble.
A minimal design on purpose
Google Cloud kept OKF small on purpose. Only one field is required. This light design is a feature, not a gap. The aim is to keep the writer and the reader apart. In other words, the tool that writes the knowledge does not need to know about the tool that reads it.
This makes OKF very flexible. It is meant to work with any cloud, any database, and any agent framework. An agent framework is the toolkit developers use to build AI agents. So a person can write a knowledge bundle, and an agent can use it. An agent can also make bundles. Then other agents or people can read them later.
Tools Google Cloud shipped to get started
To show how OKF works, Google Cloud also released a few reference implementations. These are ready-made examples. They prove the idea and help others copy it.
- A BigQuery enrichment agent that makes an OKF document for each table. BigQuery is Google’s big data warehouse. It stores and searches very large data sets.
- A static HTML visualizer. This is a simple web page. It shows the knowledge in an easy-to-read layout.
- Sample bundles built from real data. These include GA4 e-commerce data, Stack Overflow data, and Bitcoin data sets.
Google Cloud also updated its Knowledge Catalog. Now it can take in OKF and serve it to AI agents. The code and the full spec are on GitHub. Anyone can study and use them.
Key facts
| Item | Detail |
|---|---|
| Name | Open Knowledge Format (OKF) |
| Launched by | Google Cloud |
| Date announced | June 14, 2026 |
| Version | OKF v0.1 |
| File format | Markdown files with YAML frontmatter |
| Required field | “type” (the only mandatory field) |
| Optional fields | title, description, resource, tags, timestamps |
| Where to find it | GitHub (code + specification) |
FAQ
What is the Open Knowledge Format in simple words?
It is a standard way to write company knowledge. The knowledge is saved as plain Markdown files. Both AI agents and people can read these files. This makes scattered notes easy to share and reuse.
Do I have to use Google Cloud to use OKF?
No. OKF works with any cloud, any database, and any agent framework. The format is open. The files are plain text. So they are not tied to one company.
Why does the format use Markdown?
Markdown is simple, easy to read, and works almost everywhere. The files open in any editor. They look good on GitHub. Search tools can list them too. This keeps the format light and easy to start using.
How does OKF help AI agents?
It gives agents one clean place for facts instead of many messy ones. An agent can read the linked files to understand the data and the task. This helps it do jobs faster, like writing a correct SQL query.
Why it matters (especially for India / founders)
For Indian founders and small teams, this is a quiet but useful change. Many startups run lean. They cannot afford to lose time hunting for information. A format that keeps knowledge in one open shape can save real hours.
OKF is open and works with any cloud. So it lowers the risk of vendor lock-in. Vendor lock-in is when you get stuck with one provider because leaving is too hard. With plain Markdown files, a team in Bengaluru or Pune can switch tools later. They will not have to rewrite all their knowledge.
It also fits the AI agent wave that many Indian businesses are now testing. If your agents can read clean, shared knowledge, they make fewer mistakes. For a founder on a tight budget, fewer mistakes mean lower costs and faster work.
The takeaway
The Open Knowledge Format is a small idea with a big aim. It takes the messy knowledge spread across a company. Then it turns that into simple, linked files that AI agents can really use. It is still early, at version 0.1. So the format will likely grow. But the direction is clear. It points to open, plain-text knowledge that any agent can read. For teams betting on AI agents, that is a base worth watching.
Source: The Decoder
