Notion’s Token Town: 5 Rebuilds, 100+ Tools, MCP vs CLIs and the Software Factory Future — Simon Last & Sarah Sachs of Notion

Notion’s approach to building AI agents centers on a "software factory" model that prioritizes iterative prototyping, developer velocity, and deep integration with the company's existing data primitives. By treating agents as autonomous systems capable of self-verification and debugging, the team has moved away from rigid, few-shot prompting toward goal-oriented tool calling. This evolution relies on a robust evaluation framework where model behavior engineers—rather than just software developers—triage failures and refine agent performance. The platform’s strategy emphasizes "progressive disclosure" to manage tool complexity, ensuring that agents remain lightweight and permission-focused. Ultimately, Notion positions itself as the central system of record for enterprise work, where agents automate tedious bookkeeping and data capture, allowing human teams to focus on high-level collaboration and problem-solving rather than manual process management.