OKF for Claude Code: structured, portable memory your agent (and team) can read

Marco 'Gatto' Boffo — Sun, 28 Jun 2026 12:44:50 +0000

The problem: agents forget your project every session

If you pair with a coding agent, you have lived this: a new session starts and the context is gone. The agent re-discovers your auth flow, re-guesses why a decision was made, re-reads the same files to rebuild a mental model you already explained yesterday. Project knowledge — the why behind your systems, the runbooks, the "don't touch this, here's the reason" — lives scattered across wikis, code comments, and people's heads. None of it travels with the code, and none of it survives a fresh context window.

CLAUDE.md helps, but it's for standing instructions, and it gets loaded wholesale into every prompt. Auto-memory captures what an agent picked up, but it's implicit, per-agent, and not reviewed. A wiki is for humans and needs exporting. There's a gap: curated team knowledge that's structured, versioned with the code, and readable by any agent or person.

What OKF is

Open Knowledge Format is an open, vendor-neutral format (announced by the Google Cloud Data Cloud team in June 2026, Apache-2.0) that represents knowledge as a directory of markdown files with YAML frontmatter. That's the whole idea. No schema registry, no runtime, no SDK. If you can cat a file you can read it; if you can git clone a repo you can ship it.

A bundle looks like this:

.okf/
├── index.md                  # progressive disclosure (root carries okf_version)
├── log.md                    # ISO-dated change history, newest first
├── services/auth-api.md      # one concept = one file; path is its ID
├── datasets/orders-db.md
├── decisions/use-okf.md
├── runbooks/payment-failures.md
└── metrics/checkout-conversion.md

Each concept needs exactly one thing to be conformant: YAML frontmatter with a non-empty type. Everything else is optional.

---
type: Service
title: "Auth API"
description: "Issues and verifies short-lived access tokens."
resource: https://github.com/acme/auth
tags: [auth, platform]
timestamp: 2026-06-14T10:00:00Z
---

# Endpoints
| Method | Path     | Description               |
|--------|----------|---------------------------|
| `POST` | `/token` | Exchange creds for a JWT. |

Concepts link to each other with plain markdown links, which makes the bundle a graph of typed, cross-referenced knowledge — diffable, reviewable in PRs, and portable across tools because it's just text.

What okf-skills adds

OKF is the format. okf-skills is the Claude Code-native toolchain that teaches your agent to actually use it, driven by the verbatim v0.1 spec. It ships as a Claude Code plugin and as agent skills (installable on Cursor, Codex, and 20+ agents via skills.sh). Three pieces:

Skill	What it does
`/okf:okf`	Produce, maintain, and consume bundles, applying the spec and templates. Auto-triggers when a repo already has an OKF bundle.
`/okf:validate`	Deterministic §9 conformance check — a real Python checker, not an eyeball pass.
`/okf:visualize`	Render a bundle to a self-contained interactive HTML graph.

The validator and visualizer are standalone scripts too (uv run, PyYAML via PEP 723), so they work the same whether installed as a plugin or as skills. The visualizer output is a single viz.html — concepts as nodes, links as edges, a wiki-style detail panel with rendered markdown and "Links to / Cited by" backlinks. No backend; nothing leaves the page. There's a live demo of a real bundle, and the repo documents itself in OKF, with CI validating that bundle on every push.

Honest scope

A few things worth being straight about:

OKF is curation, not magic. It's a place to put knowledge deliberately, not a system that captures it automatically. Someone (you, or the agent on request) still writes the concepts. The payoff is durability and portability, not zero effort.
It's complementary, not a replacement. Use CLAUDE.md for how to behave, auto-memory for what the agent picked up, and an OKF bundle for what the team knows. They stack.
No hooks by design. The plugin won't silently rewrite your repo. Automatic upkeep is opt-in (see below).
The format spec is Google's, vendored verbatim with attribution; this plugin's own code is MIT © Marco Boffo.

Install and try it

As a Claude Code plugin:

/plugin marketplace add scaccogatto/okf-skills
/plugin install okf@scaccogatto

As agent skills (Claude Code, Cursor, Codex, 20+ agents):

npx skills add scaccogatto/okf-skills

Then capture, validate, and visualize:

/okf:okf produce .okf          # or just ask: "document the auth service in OKF"
/okf:validate .okf --strict
/okf:visualize .okf            # opens a shareable viz.html

Every concept gets a deep link — viz.html#services/auth-api loads the graph with that concept already selected.

To turn on automatic upkeep (consult .okf/ before tasks, write knowledge back after changes), paste templates/CLAUDE-okf.md into your project's CLAUDE.md. It's soft mode and entirely opt-in. The scripts need uv (or python3 + pyyaml).

Why bother

The pitch is small and concrete: give your project's knowledge one durable, diffable, portable home, versioned next to the code it describes — instead of re-explaining it to a fresh agent every morning. It's plain markdown, so worst case you've written some good docs your team can read. Best case your agent stops forgetting.

Repo, demo, and spec: github.com/scaccogatto/okf-skills. Issues and PRs welcome.