Where OKF Could Go

OKF is a young draft, so its future is a question, not a plan. This is our read of where it could go, held honestly against what it does not yet do.

Where the format is heading

A few directions look plausible. Richer provenance and verification could become a norm, not just a citation habit. Agent-native access, through interfaces like MCP, could become a first-class way to distribute a bundle. Conventions for discovery and trust could emerge, alongside an ecosystem of tools and bundles that reaches beyond any single knowledge base.

What v0.2 closed, and what is still open

Credibility here depends on naming the gaps plainly. The format is young, and it shows.

Provenance was the gap we named first, and v0.2 closed most of it. The format now carries a sources block in frontmatter, a verified record naming who confirmed the content against those sources, and trust tiers a consumer derives from that record (OKF SPEC, v0.2). A reader can now ask a bundle what has been checked, and by whom, through the format itself.

What is still open is the harder half. The format records that a check happened; it cannot tell you the check was any good. "Verified" remains an assertion by the publisher, and a reader has no way to weigh one publisher's assertion against another's.

There is no fixed concept taxonomy. That is a deliberate choice. It keeps authoring flexible, but it costs you consistency across bundles and makes discovery harder.

There is no standard for discovery or registry. You have no defined way to find bundles, or to know that a given bundle is authoritative. Versioning and governance are also early, as you would expect from a draft.

What would have to be true

For the trajectory above to hold, several conditions would need to fall into place.

Provenance has grown past a convention into something a machine can read — that happened faster than we expected. Turning it into something a machine can check is the remaining step, and it needs adoption rather than more spec: enough publishers recording verification the same way that a reader can compare across bundles.

Discovery would need a real standard, so bundles can be found and ranked by authority. Governance would have to mature enough to carry breaking changes without fracturing the ecosystem. The versioning plan already signals intent here: minor bumps for backward-compatible additions, major bumps for breaking changes. Adoption is the harder part. A format becomes a standard when tools and authors outside its origin choose to build on it.

Our read

This is our view, not a forecast and not the maintainers' roadmap. We think the direction is sound. The gaps are the honest kind: known, named, and the sort a draft standard is expected to close over time. The open question is not the design. It is whether enough people build on it to make discovery, trust, and governance real rather than planned.

Sources

• Open Knowledge Format specification — OKF SPEC.md (GoogleCloudPlatform/knowledge-catalog), v0.2. verified: A list of records of who or what has confirmed the content against its sources; No verified key means unverified; verified by non-human actors only means machine-confirmed; verified by a human: actor means human-reviewed. View source · verified 2026-08-25 · primary