Open Knowledge Format: The Markdown Standard for AI Agents
AI agents are only as smart as the context you give them.
Most teams face the same wall. You have a capable model and a solid framework. But the agent knows nothing about your organization. It does not know your database tables, your metric formulas, or your incident runbooks.
This knowledge is scattered across Notion, Slack, and wikis. To give it to an agent, you usually build custom integrations for every single source.
The Open Knowledge Format (OKF) changes this.
OKF is a vendor-neutral specification. It packages your knowledge as plain markdown files. Any agent can read it without custom SDKs or proprietary tools.
How it works: An OKF bundle is a directory of markdown files. Each file represents a concept like a table, a metric, or an API.
Why this is effective:
- Identity through file paths: The file path is the ID. No database or registry is needed.
- Knowledge graphs via links: Markdown links create relationships between concepts. This turns a folder into a navigable graph.
- No tooling required: The file system is the API. You do not need new software to use it.
How to use it:
- Data teams: Document tables and metrics so agents write accurate SQL.
- SRE teams: Write runbooks so agents can diagnose incidents.
- Platform teams: Ship API definitions so agents understand endpoints.
OKF is not a replacement for RAG or MCP.
- Use OKF for stable, curated knowledge like schemas and runbooks.
- Use RAG for large, unstructured document archives.
- Use MCP for live data and tool connections.
This format is part of the Apache 2.0 license. It is cheap to ship and prevents vendor lock-in. Your knowledge lives in your git repository and stays readable by any tool.
Start small. Pick one team with messy documentation. Write three OKF files and point an agent at them.
Optional learning community: https://t.me/GyaanSetuAi