The Context team owns the integration service (the shared substrate for connecting to third-party systems: OAuth flows and token lifecycle, API key management, MCP servers and clients, credential storage and rotation, webhook ingestion, and per-provider rate limiting and backoff); the integration catalog itself (CMS platforms, analytics and search consoles, CRMs, document stores, support desks, and commerce systems, each with its own auth model, pagination quirks, and failure modes); knowledge base storage and retrieval (ingestion and chunking, embeddings, hybrid keyword and vector search, reranking, freshness, and strict tenant isolation across a store handling billions of documents); data synchronization (incremental sync, backfills, change detection, deduplication, conflict resolution, and the observability that tells us a customer's data is current before they notice it isn't); and context management (the knowledge graph that resolves brands, products, competitors, topics, and pages into a coherent model of a customer's world, plus the long-term memory service that lets agents accumulate and reuse what they learn). Drive technical decisions on hard problems: retrieval quality and how we measure it, graph and memory schemas that survive contact with real customer data, sync architectures that stay correct under partial failure, and an integration framework that makes the tenth connector cheaper to build than the first.