Open Knowledge Format (OKF)
The Open Knowledge Format is an open standard for knowledge that AI agents can read: plain Markdown with a little YAML front-matter, organized in folders, with links between documents. Hjarni is built on the same primitives, and it can export your knowledge as an OKF bundle in one click.
What OKF is
The Open Knowledge Format (OKF) is an open standard published by Google Cloud in June 2026. It describes how to write down knowledge so that AI agents, and the people building them, can read it without a custom integration for every source.
It is deliberately small. OKF is plain Markdown files, with a little YAML front-matter at the top, organized in folders, with ordinary Markdown links between them. No database, no SDK, no runtime. You can read it in any editor, render it on GitHub, and ship it as a folder or a ZIP.
It is a format, not a platform. The tool that produces the knowledge and the tool that consumes it can change independently, because both speak the same files. That is the same bet Hjarni made: your knowledge should be plain Markdown you own, not rows in someone's database.
The format
Each concept is one Markdown document. The front-matter carries a small set of fields, and the folder layout plus the links do the rest.
Front-matter
A type (the only required field), plus optional title, description, resource, tags, and timestamp.
Folders, index.md, and log.md
Documents live in folders. Each folder can hold an index.md that acts as a map, and the bundle can carry a log.md that records what changed.
Links as a graph
Links between documents are plain Markdown links. Followed together, they form a graph an agent can walk from one note to the next.
The specification and reference tools are on GitHub. OKF v0.1 is explicitly a starting point built to grow.
Export your knowledge as OKF
Hjarni can package your notes as an OKF bundle. You get a ZIP with one Markdown file per note, an index.md for every folder, a root index.md and a log.md, and your wiki-links rewritten as relative Markdown links so the graph stays navigable outside Hjarni.
- Open your folders page, a single folder, or a team.
- Click Export, then choose Open Knowledge Format.
- Download the ZIP. It contains your notes as OKF documents in folders that mirror your structure.
Prefer Hjarni's own front-matter instead? Pick Markdown from the same menu. See export and data ownership.
What lands in the bundle:
One .md file per note, with type: Note and OKF front-matter (title, timestamp, description, resource, tags).
An index.md per folder (type: Collection), carrying the folder's description and AI instructions plus links to its notes and sub-folders.
A root index.md linking every top-level folder, and a log.md change history built from your note versions.
Wiki-links between notes become relative Markdown links between files.
File attachments are included alongside their note in an attachments folder.
The bundle round-trips. Drop the ZIP into import and your folders, note links, summaries, and source URLs come back, while the index.md and log.md navigation files are skipped.
How OKF maps to Hjarni
Hjarni's model lines up with OKF almost one to one. That is why the export is a clean translation rather than a rebuild.
A note (Markdown body plus metadata)
Note for a note, Collection for a folder index, Log for the change history
Note or folder title
When the note was created (with the last update alongside it)
Note summary
Note source URL
Tags
A container (folder), with its description and AI instructions
A change history built from note versions
Wiki-links and backlinks between notes
Why it matters
Knowledge that lives in one vendor's box is only as useful as that vendor's AI. Knowledge written as plain, structured, linked files can be read by any agent you connect, today and in years. That is the difference between renting your memory and owning it.
An OKF bundle is portable by design. Hand it to another OKF-aware tool, drop it in a git repo, or point your own scripts at it. Nothing is locked into a proprietary shape, and a major cloud publishing the same format is a strong signal that plain Markdown is where agent-readable knowledge is heading.
Most days you will not need to export at all. With Hjarni's built-in MCP server, Claude and ChatGPT read and write the same notes live. OKF is there for the day you want your knowledge to travel.
Common questions
FAQ
What is the Open Knowledge Format?
An open standard, published by Google Cloud in June 2026, for knowledge that AI agents can read. It is plain Markdown with a little YAML front-matter, organized in folders, with links between documents forming a graph. A folder can hold an index.md as a map, and the bundle a log.md for history. The only required front-matter field is type.
Does Hjarni export the Open Knowledge Format?
Yes. Open the Export menu on your personal space, a folder, or a team and choose Open Knowledge Format. Hjarni builds an OKF bundle: one Markdown file per note (type: Note), an index.md for every folder (type: Collection), a root index.md and a log.md change history, with wiki-links rewritten as relative Markdown links. It downloads as a ZIP you own.
How is OKF different from the normal Markdown export?
The normal export uses Hjarni's own front-matter (title, summary, tags, source_url). The OKF export makes each document conformant: the required type field, OKF field names (title, description, resource, tags, timestamp), a per-folder index.md, a root index.md and log.md, and wiki-links converted to relative links so any OKF-aware tool can follow the graph.
Who created the Open Knowledge Format?
Google Cloud published OKF v0.1 in June 2026 as an open, vendor-neutral specification. There is no SDK or platform to adopt. It is a format, not a product, so the bundle works with any tool that reads Markdown.
Can ChatGPT and Claude read OKF files?
Yes. OKF is Markdown, so any AI can read an exported bundle. You usually do not need to: with Hjarni's built-in MCP server, Claude and ChatGPT read and write the same notes live, instead of you exporting files by hand.
Give your AI a memory. Free.
Connect Claude or ChatGPT to notes they can actually read and write.
Give your AI a memory. Free.