Google Cloud has launched the Open Knowledge Format (OKF), a new specification that standardizes knowledge as Markdown files to enhance AI agent interoperability. The format enables knowledge to be portable across systems and structured as a directory of Markdown files with YAML frontmatter. OKF v0.1 includes a required field for 'type' and optional fields such as title, description, and tags, allowing for flexible data modeling. Concepts are linked through standard Markdown links, forming a knowledge graph that can be read in any editor, rendered on GitHub, and indexed by search tools.
The OKF aims to address the challenge of fragmented knowledge, which slows down AI agents. According to Google Cloud, knowledge is often scattered across metadata catalogs, wikis, code comments, and the minds of engineers. AI agents must piece together information from various sources to perform tasks like writing SQL queries. The company states that current solutions are custom-built and not interoperable, leading to knowledge being locked within specific systems. OKF seeks to close this gap by offering a minimal, portable format that can be used across different cloud providers, databases, and agent frameworks.
Google Cloud is also releasing several reference implementations alongside the OKF specification. These include an enrichment agent that crawls BigQuery datasets to create OKF documents, a static HTML visualizer, and sample bundles for GA4 e-commerce, Stack Overflow, and Bitcoin datasets. The Knowledge Catalog has been updated to ingest OKF and serve it to agents. The spec and code are available on GitHub, with the Knowledge Catalog integration documented separately.
Source: thedecoder
