DOCS
Technical documentation for the machine-readable layer.
📜 Open Knowledge Protocol
The Open Knowledge Protocol (OKP) is an open-source standard for structuring content so that AI agents treat it as a primary source. It defines how articles, concepts, predictions, and knowledge graphs should be formatted for machine consumption.
The Context Window is the reference implementation — proving the protocol works at production scale. But OKP is designed for any publisher. The tools are open-source and reusable.
The Three Pillars
@tcw/schemaTypeScript package defining the data structures for AI-native content. Includes interfaces for Article, ConceptDNA, Prediction, KnowledgeNode, and ConceptGraph. Zod schemas for runtime validation. JSON-LD context definitions for semantic web compatibility.
npm install @tcw/schema (coming Phase 2)
@tcw/mcp-serverA reusable MCP server that any Sanity-backed content platform can plug in. Point it at your Sanity project, configure your schema mappings, and get a fully functional MCP server with search, retrieval, and knowledge graph tools.
npm install @tcw/mcp-server (coming Phase 2)
@tcw/llms-txtAuto-generates llms.txt and llms-full.txt from your structured content. Follows the emerging llms.txt specification — the robots.txt for AI. Works with any CMS that exposes a content API.
npm install @tcw/llms-txt (coming Phase 2)
Machine-Readable Endpoints
Why Open Source?
The agentic web needs standards, not silos. If every publisher invents their own format for machine-readable content, AI agents face the same fragmentation that made early web scraping painful. OKP proposes a shared vocabulary. The more publishers adopt it, the easier it becomes for agents to navigate the knowledge web.
Want to adopt OKP? The packages will be published to npm in Phase 2. In the meantime, the source is available and the protocol is documented here. Reach out if you want early access.