# Context Hub Context Hub is a CLI tool and knowledge management system created by [[Andrew Ng]] that helps AI coding agents access curated, versioned API documentation and skills. It is distributed as an npm package (`@aisuite/chub`) under the MIT license. It addresses a core limitation of AI coding agents: they hallucinate APIs and forget what they learn in a session. Context Hub provides persistent, curated documentation that agents can reference and improve over time. ## How it works - Search available documentation: `chub search "stripe payments"` - Fetch language-specific docs: `chub get stripe/api --lang js` - Agents read curated docs and write correct code Content consists of plain markdown files with YAML frontmatter, maintained in the repository for transparency. Users can inspect exactly what agents read and contribute improvements via pull requests. ## Self-improving loop Agents can annotate gaps locally (`chub annotate stripe/api "Needs raw body..."`), and annotations persist across sessions. Feedback (upvote/downvote) flows back to content authors, so community documentation improves incrementally. ## Key features - Multiple language variants (Python, JavaScript) - Incremental fetching with `--file` and `--full` options - Version control for documentation - Token-efficient selective content retrieval ## References - https://github.com/andrewyng/context-hub ## Related - [[Artificial Intelligence (AI)]] - [[Large Language Models (LLMs)]] - [[Andrew Ng]]